CPUC Workshop - 2021-02-22
No chapters available yet
Legend
Transcript markers
- text Sentence being played
- text Search result
- text Intervention cited by the link you opened
Click a sentence to play the video from that moment. Select text to quote it with the speaker, timestamp and link.
More options
California Public Utilities Commission
Follow its sittingsSpeaker 2
We're getting some background noise coming in. So please mute mute yourself. So please only chat with the host directly if you have a question about a technical Webex issue. Otherwise, questions directed to the host will likely not be addressed in any of our panel discussions. If you wish to ask questions to the presenters, please select all panelists in the chat function.
Someone from the wildfire safety division will be monitoring all the posts directed to all panelists. Again, please do not use the Q and A feature, which comes in under the chat. That is, we're not going to be monitoring the Q and A feature, we're only going be monitoring the chat today, and we will route the questions to the appropriate panelists or just present them to the panelists themselves. Just a final note is that in addition to the panelists, the utility representatives, we've also asked several stakeholder organizations to serve as panelists, so they will be able to have an opportunity to ask questions directly of the utilities. Can we go to the next slide, please?
Okay, so as an example, when we go into the risk mapping and assessment portion, stakeholders that are designated as panelists may ask questions. So in order to ensure an orderly discussion, we ask that the panelists please use the raise hand function in order to ask questions. The raise hand function is found in the participant list, so if you open the participant list and get that on the right hand side of your screen, it's next to your name. The moderator will coordinate those stakeholder panelists' questions in the order the hands are raised. Once the stakeholder panelists ask their questions, we ask that you lower your hands by clicking the same button to clear the queue.
For everyone else who does not have panelist status, again, please use the chat function as described.
Speaker 2
of questions received, we may not be able to get through all of them, but we will definitely give it our best and try to get through them. So before I hand it over for our first panel discussion, which we're actually a little bit ahead of time, Are there any questions that anyone has? If so, free to please put those into the chat and I can do my best to answer them. Okay, so I'm not seeing any questions. So with that, I'm going to hand it over to Alan Wu, and he's going to moderate the first portion of our workshop.
Thank you again, everyone, for joining. Looking forward to a fruitful two days.
Speaker 4
Thank you, Melissa. Hi, everyone. My name is Alan Wu, and I'm a wildfire safety analyst with the Mitigation Branch of the Wildfire Safety Division. I'll be your moderator for the upcoming session. Wildfires continue to be a big threat to the environmental, economical, and social well-being for the State Of California, and wildfire risk models will play an important part of this battle by being the threat that holds the entire wildfire mitigation plans together.
Risk models should show the utilities where their risks lie now, where their risks will be in the future, and also act as a guiding compass to pinpoint where resources should be distributed. For the next hour, we will focus on risk assessment, mapping, and resource allocation methodology, and listen to the three utility companies explain their wildfire risk models and how they're incorporated into the 2021 WMPs. The utilities will present in the order of PG and E, SCE, and SDG and E. Members of the audience and panel, please remember to reserve all questions until the Q and A portions of today's schedule. PG and E, we welcome you first.
Speaker 5
Good morning, everyone. Just a quick sound video chat. I'll take that as affirmative. I'll introduce myself first of all. My name is Paul McGregor.
I'm the Director of Electric Operations, Risk Management and Analytics at PG and E. I'm new to PG and E, having joined them in 2021, so I think it's important that I take a minute just to explain my background. Previously I led the enterprise risk management function at Southern California Edison 2019 to 2020, where I developed the wildfire risk model for SCE as well as participated in the general rate case for the 2021 filing. Along the way, we developed the wildfire mitigation plan for 2020 and I initiated as part of the team for the 2021 filing. Prior to that, I served as a consultant to the electric utilities industry, primarily California, 2017 through 2019, where I assisted PG and E with their 2017 RAMP filing, SCE in their 2018 Ramp filing, and PG and E with their 2020 General Rate Case.
So being in and around California, I'm very familiar with wildfire. Prior to that, I've spent about twenty five years in the electric utilities industry in a number of operational and financial and risk management roles. So that's a little bit about me. We'll jump to the next page. I have the pleasure this morning, folks, of introducing you to PG and E's wildfire mitigation plan for year 2021.
And we can go to the next page as well. And then I've just got a couple of pages to lay the landscape. Basically PG and E has a 70,000 square mile service territory. Alan, could you go to the next page, please? It isn't turning online.
About half of our service territory lies within the High Fire Threat Districts, Tiers two and Tiers three, and nearly one third of our overhead lines are in the high fire threat district areas. And we power our customers and communities. As you can see, about 10% of our customer communities are in those high fire threat districts as well. In recent years, we've seen high temperatures, extreme dryness and very high winds in our service territory, again, across California, also in PG and E. 2020 was another year of unprecedented fires.
Five of the six largest fires recorded, five of them were in PG and E's service territory. And then for 2021 we've kicked off with a red flag warming and a PSPS event in January 2021. So the environment doesn't seem to be letting up there. If we can jump to the next page, PG and E's 2021 wildfire mitigation plan continues many of the actions from the 2019 and 2020 WMPs. It reflects our evolution, embracing and advancing of technologies both in risk assessment and in risk mitigation.
We apply a number of the lessons learned from the 2020 WMP process and incorporating feedback from the Welfare Safety Division PG and E has a federal monitor and several other industry partners. The welfare mitigation plan has basically three goals: the reduction of wildfire ignition risk by inspections, vegetation management system hardening SPS where required. The enhancement of situational awareness with weather stations, cameras and advanced meteorology. And finally, to reduce the impact of PSPS on our customers by reducing the number of customers impacted, the duration of those customers impacted and improved timeliness and accuracy of information to communicate to those customers and communities. All of this is supported by our wildfire risk models, and we have significantly enhanced those.
And that's really what we want to talk about today. So if I could ask you to advance, it's two pages, Alan, actually, because there's a slide in between. This gives you an overview of our evolution, and as I explained through my background, I've actually had the experience of working with PG and E through a number of these stages along the way. We started in 2017 with our first Ramp Report, which was a first generation set of models using probabilistic modelling and Monte Carlo simulation, the implementation of our risk spend efficiency metric. And again, first mover, first step forward.
In 2019 for our WMP, we developed the Community Wildfire Safety Programme again implemented first of a kind probabilistic prioritization models using fire spread modelling. For 2020, again continuation of these programs as we moved to the Tecnosilva wildfire propagation and consequence modeling. And then in 2020 we filed our second ramp filing, which again was a second generation set of ramp models. This time we moved to Python interface and we enhanced our multi attribute value function in accordance with the 2018 S map settlement agreement for 2021. I'm offering you our next generation of risk models where we adopt the maximum entropy modeling approach as well as marrying that up to the techno Silva consequence and propagation modeling.
So it's been quite a journey. I think the words of David so called captures this best where I would see I'm pleased with progress but not satisfied. So we have a long way to go. To the next page, please. Our 2021 wildfire risk modeling approach very much supports and aligns with our ramp.
It also endorses the probability of ignition component upfront or likelihood of risk event lower as we call it, as well as that core, the consequence of a risk event. The model is driven predominantly by our vegetation ignition model and our equipment ignition model coming together and combining with the techno Silva consequence model to produce our output. Our equation is simple wildfire risk equals ignition probability multiplied by wildfire consequence. As we discussed before, mitigation programs focus on system hardening, enhanced vegetation inspection and repairs, and each of these are prioritized using the outputs of this risk model. Moving on to the next page, I would like to talk a little bit about our enhancements from previous models.
So there's a lot going on in this slide, so I'll take a second. When we started this journey, we used our 2019 risk models which were built at the tail end of 2018 using a regression type approach, and we married that up with the reacts propagation and consequence. And we also had some other features in there to assume egress. For 2021, we've made a number of enhancements, the first major one being replacing the regression based vegetation ignition likelihood model with the maximum entropy machine learning ignition probability model and an expanded covariant pool. We did a very similar thing with our ignition, sorry, our equipment ignition model, which focuses on conductor failures.
Again, it utilises the maximum entropy machine learning approach and also has a wider span of covariate pool. These are married together with our TechnoSilva consequence model. So Tekna Silva uses a more advanced set of data layers than we previously used in WEAX, is an updated fuels data using LANDFIRE 2016, which is enhanced for disturbances through 2020. Disturbances mean fire events. It also sits on top of updated population information and updated structures information.
When we look at consequence coming out of the consequence model, we're looking at acres, we're looking at buildings impacted and population impacted and also a feature called fire behavior index, which gives us an estimation of where that fire is going to go and at what kind of intensity. If I could jump to the next page, I want to talk a little bit about this maximum entropy approach and I think it's important to set out what this is exactly. It's basically a logistic regression. Alan, I believe there's a delay if you could maybe move that slide forward along one. Thank you.
That's great. So what does maximum entropy do? Maximum entropy is an algorithm that was developed on a logistic regression perspective, and it's particularly good for identifying where a species may be found based on the habitat. So how does that translate? Quite simply, what we're looking for is that.
Where can I expect something to occur based on the covariance or variables around the area? So what I'm looking at is I'm looking at a set of asset variables, namely conductor size, conductor age, are there splices in that conductor etc. So specific to the assets themselves. And then we're looking at a number of environmental variables also. So when I say environmental, I'm talking about vegetation, precipitation, the dryness, the amount of fuel.
Is it ground fuels? Is it ladder fuels? Or is it high burning crown fuels? And finally, the weather covariance. So we're looking at the weather conditions, we're looking at wind, we're looking at precipitation, we're looking at temperature and also applying a temporal perspective to this.
In other words, when will these covariates come together and produce the ignition effect? So that's a basic introduction. We try to keep this simple and basically seeing where can we expect this to occur. So once we generate our probability of ignition, I can jump to the next stage and show you the progression. One second, thanks sir.
So again, we developed the probability of ignition using this maximum entropy approach. And from that, we marry that to the techno Silva propagation and consequence model. As I discussed, it gives us our outputs in natural units. They are acres, structures and people as well as fire behaviour index. We also convert this to what PG and E calls its core MAVF.
So that's a consequence of risk event multi attribute value function. And we use those particular points for internal discussion when we're ranking this. So we have ignition probability multiplied by consequence coming together, providing us with our risk score. And we can see this in a geographical interface. So just moving on to the next slide.
Again,
Speaker 5
want to emphasise how we operationalise this model because we've got a lot of data. It's a machine learning interface, so it's lots and lots of CSV files. And then we're overlaying that on top of a GIS dataset. So we take the outputs of the risk model and we have to consider Additional factors and updated language. Risk model were all built in the twenty fifteen to twenty nineteen set of data.
2015 to 2018 was our training data set and 2019 was our testing or verification data set. We apply some twenty twenty findings, whether they be in our updated LiDAR on strike trees within the higher priority rec districts or from our public safety specialist expertise in the area, knowing these particular areas, knowing the patterns, egress, the terrain, etc. A lot of these people are ex fire service, local and statewide. So it gives us a good insight to fire behaviour in that particular area. And then we want to consider additional system hardening projects that have gone on and particularly fire builds where we've seen a fire in recent years.
And also to consider PSPS de energisation impacts from our 2019 and 2020 seasons. These come together, they inform our work plans for vegetation management, for system hardening and also our metrics which are in our wildfire mitigation plan and also part of the company's compensation plan. So to give you guys an example here, I'll take you through one example on the next page, please. This is an example of a circuit that we call Keswick 1101. It's in the Whiskeytown area of Shasta.
So anybody that knows that area, it's a very beautiful area, large recreational area, a beautiful lake surrounded by trees. So on this particular circuit, we've got the risk model identifies this as a high priority circuit based on the COVID rates that we talked about, the age of conductor, the size of conductor, the proximity to drive vegetation, to strike trees and applying weather variables also. So this particular segment, so it's called, we call it a circuit protection zone, which is an isolatable section of a circuit, simply a circuit segment. This pops out as one of our top 50 miles. It's about 6.6 miles in length.
When we build our risk model, we're building this at 100 meter by 100 meter grid squares, and then we aggregate those grid square pixels to the particular circuit segment or circuit protection zone. So the score for this one in our language is 48.8 MAVF units. Our total MAVF units just coincidentally are about 25,000, which is a similar amount of distribution overhead lines we have in the High Fire Threat Districts. So it's 49 points effectively. And the average risk score on this particular segment is about 1.25 units.
So it's relatively high risk. And again, we consider this circuit for various system hardening activities. We consider undergrounding, we consider traditional hardening such as covered conductor, And we also consider moving parts of this circuit or removing parts of this circuit if we can do so. For this particular example, I believe the final solution was a hybrid of both undergrounding and covered conductor system hardening type activity. So moving on to the final page from my presentation is really what are we doing next?
What does 2021 look like for PG and E? So for me, there's a lot of work in advancing and evolving our wildfire mitigation programs and also understanding our risk. So for our nomenclature, we call this current model our 2021 wildfire distribution risk model, and we will enhance some of the capabilities for that to build out our 2022 wildfire distribution risk model. Some of the things we'll be doing there is adding additional asset classes, including transformers and poles, and also breaking down that granularity to help us look at the RSE at a localized circuit segment type level. Also in development is the 2022 wildfire transmission risk model.
Again, slightly different set of data will be used for that one. We're looking at different asset classes. There are lesser ignitions at the transmission voltage class, so we need to be specific when we're training our models on what to look for. And finally, we're building out a future state PSPS consequence model. PGD has built a consequence model for the enterprise or portfolio level.
What we're doing now for 2021 is building that out to a circuit level to allow us to forecast a probability of de energisation and the potential impact on the customers and communities within that particular segment of the circuit. So that ends my presentation as far as the formal. I know we have a Q and A session. I see there's a couple of questions in the chat. I'm happy to answer those now but I'll have to look at the moderator if you want to hold those, do those now or move those on to the Q and A session for later in the day.
Speaker 4
Yep, thank you so much, Paul. We're going to reserve all the questions until the Q and A section, so we'll get to them there. Yeah, thank you so much for the presentation and we have Southern California Edison coming up next. Thank you.
Speaker 7
Okay, good morning everyone. My name is Robert Lemoyne. I'm Director of Enterprise Risk Management for Southern California Edison, and can we go to the next slide, please? I want to first thank the Commission and the Wildfire Safety Division staff for inviting us today, here today, and also recognize Paul McGregor as well for setting the stage for us. Really appreciate that, Paul.
I'll be taking the first two slides that you'll see in this presentation, and I'll be handing it over to Joe Gazetta, our director in T and D, who is responsible for the wildfire risk model. Could we go to the next slide please? Okay, so what I want to talk to you about today is how we have evolved our risk modeling capabilities since we began the journey back in 2018 with our grid safety and resiliency plan back in 2018. We've evolved our risk models significantly since then. Back in 2018 when we first started this journey, process was to take a look at the exposure of our lines in high fire risk areas.
So this was a very early stage portion of determining risk, and the most important characteristic in our GSRP plan was how much mile of conductor was contained in the defined high fire threat districts, tier two and tier three. That was married up with some other probability of ignition models that allowed us to determine where we wanted to target our initial grid hardening efforts. Those grid hardening efforts led to significant work that was performed throughout 2019 and 2020, and then we continue to evolve those programs. Starting in 2019, we evolved our capabilities to first
Speaker 7
looking at a much better consequence model that REAC's engineering provided to us in early twenty nineteen. That consequence model was a static model that looked at fire spread based on location throughout our service territory in the high fire threat districts. It modeled structures impacted, population impacted, and did so using worst weather days from the previous twenty year data set. That model, married up with our probability of ignition model, allowed us to again begin to target our grid hardening efforts on a much more granular level. Starting this year, what we're able to do was take those models that were based on the tiers, high fire threat tiers two and three, and begin to tranche our consequence and probability ignition down to the circuit segment and asset level.
Speaker 7
by first contracting with Technosilva, the same company that was mentioned in the presentation, which gave us real time capabilities to and much more updated data for consequence. Taking that down to a circuit and segment level allows us to really model the areas of our service territory that are the highest risk and to rank those areas for grid hardening efforts. In addition, this past year, we developed an approach to estimating TSPS risk. If we can go to the next slide, I'll go over how we did that. Okay, so in mid-twenty twenty, we began developing what we call our MARS two point zero.
It's the extension of our multi attribute value function from our 2018 ramp proceeding and to conform with the 2018 SMAT settlement. In evolving to MARS two point zero, again, one of the main features of this was to be able to tranche our work from what in the last year's WMP was a very high level, a system level framework, down to a circuit and segment level framework. So we did that with two different what we call risk stacks. The first on the left on this slide is our wildfire risk stack. What this does is it takes the output from our wildfire risk model, the probability of ignition and the consequence provided by Technosylva, and translate that into a multi attribute risk score using the dimensions safety, reliability and financial.
Those are calculated using the outputs in the Tecnosilva model based on population impacted by the fire, the amount of reliability impacts, and financial impact based on amount of structures impacted. That gives us one score for each location for the wildfire risk score. We've also developed a second risk stack for public safety power shutoff. That is a combination of probability of de energization with the same dimensions, safety, reliability and financial. The probability of de energization is calculated looking back, doing a back cast of ten years look back to determine which circuits and how long those circuits would be de energized based on current approaches and the weather that occurred back in those timeframes.
On the consequence side, safety, reliability and financial are calculated based on the number of customers that are impacted by the de energization event, the length of that de energization and a framework for understanding the financial impacts to customers from a PSPS event. Doing it this way allows us to look at a combination of both the wildfire and the PFPS risk scores, allows us to add them up, allows us to provide mitigation effectiveness against these different dimensions so that we can understand how something like cover conductor can reduce both wildfire risk and the risk of PFPS to customers. By doing that, we can understand where it's best to deploy those mitigations. So what I'd like to close with before I send it over to Joe is that most of what I'm talking about, what we're talking about today is covered in our wildfire mitigation plan, majority of it is in Chapter four. I encourage everyone to take a look at that where we describe in detail how we come up with these metrics, and we'll be happy to answer questions at the end of this presentation.
With that, want to advance the next slide, I'll turn it over to Joe.
Speaker 6
Thanks, Robert. Hi, I'm Joe Goyasweta, Director of Advanced Analytics and Process Improvement here at Southern California Edison. Good morning. So Robert described these two stacks, and I'll go into them a little bit more detail so you understand how they're created. When we talk about our wildfire risk reduction model, we actually describe it as a framework because it's a collection of many, many models.
We built it that way intentionally so that we'd have a lot of flexibility to analyze problems in a lot of different ways. Both the wildfire and the PSPS risk components of the wildfire risk reduction model follow the same bowtie approach. And I'll walk you through each of these stacks: the wildfire stack, which is the bow tie at the top, and then the PSPS stack, which is the bow tie at the bottom. So, we start with the upper left box, probability of ignition, we chose to take a machine learning approach at the asset level for building our models, mainly because we wanted enough granularity to understand each potential mitigation in as much detail as possible. And we also chose to focus on outages as a proxy for ignitions rather than using ignitions themselves.
And the main reason is because, fortunately, we don't have a lot of ignitions, at least not the number of ignitions that you'd really need to build a robust machine learning model. Any of these big data models really require thousands and thousands of data points. So, the approach we took was to look at our outage data and look at the categories of outages, the classifications of outages, and identify which type of outages could have led to a spark. So, certainly, like a wire down could lead to a spark, a transformer could lead to a spark. And we chose that, which gave us upwards of 100,000 outages in our database, I think over 85,000 outages in our database, to build these machine learning models from, which now gave us enough data to really hone in on these.
So like I said, we built them at the asset level. We started with the conductor model and built conductor at the segment level, probability of ignitions for any kind of conductor failures. Then we moved on to capacitors, transformers, switches, and so for each individual asset has its own machine learning model to predict failures. Then we created a set of contact with foreign objects models. Those as well show up in our outage data.
So we know if we had a balloon outage or a vegetation related outage or a car hit pole, for example. So we built individual machine learning models for each of these different types of contacts with foreign objects: vegetation, animal, balloon vehicles. And then from that, the way we approach it is we run each of the models at the asset location. So typically, since we're really dealing with mostly overhead areas, we're really talking about at the pole level, we call them a FLOC, a functional location area. And what it does is we allow, we calculate the probability of admission at each pole for everything outside of conductor, which we do at the segment level, and then we aggregate those probabilities.
So a pole that has multiple pieces of equipment will aggregate those probabilities, because each one of those assets has a potential of creating a spark. From there, we then multiply that by our consequence score. As Robert mentioned, one of the advancements we made last year was moving from our REACTS consequence scores to the new Technosova consequence scores. That gives us our risk score for wildfire. And then, like I said, depending on the analysis we're doing or the problem we're trying to solve, we can either use the full wildfire component or we can use an individual wildfire component.
So for example, when we're scoping our covered conductor, the way we'll approach that is by looking at the total segment risk, creating a risk curve similar to what you saw in the PG and E presentation. And then we will take out the tranches from the highest risk to lowest risk based on how much scope we want to target. So for example, when we're ready to go after the next 500 miles, we go down that curve to the next 500 miles worth of segments that are the highest risk, that's what goes over to engineering. And then engineering will sort through the operational components of it to figure out the best way to operationally get it constructed. We could do that for various other things.
So for example, for our inspection programs, the inspection programs will look at both the probability of ignition, along with the consequence as two different dynamics, because what we've been targeting with our inspection programs is making sure that our accelerated or enhanced inspections, accelerated inspections are looking at both the highest consequence and the highest probabilities. For vegetation, vegetation may only look at consequence and prioritize their work based on consequence where it's appropriate. One of the main things that the reason we wanted this extra flexibility is because we always have to balance the operational efficiencies that you gain by approaching things by area versus things where you want to go after in order the highest risk to the lowest risk. Because there's always that time component of how much risk can you buy down if you work more efficiently versus purely based on risk. On
Speaker 6
bow tie, you see our probability, I mean, PSPS component of the wildfire risk reduction model. And as Robert mentioned, essentially what we did to establish our probability of de energization was to go back through ten years of weather data, both wind and FPI data, and then calculate based on our current protocols for PSPS de energizations, each time we would have, if the PSPS program was in place in the past, how many times we would have or what the frequency of the de energizations would have been for each isolatable segment. That basically then gives us a probability, assuming that those same patterns are the same moving forward, gives us our probability of future PSPS events. Those are then connected to PSPS consequences, which use components of the some data from the techno SOVA and then implemented through our MARS risk framework. Both of these streams together, that's why we reason we call them stacks, is that we can add the wildfire risk along with the PSPS risk to calculate a total risk figure, Or we can look at just wildfire risk or just PSPS risk, depending on what the problem is to be solved.
It also gives us a sense for how does PSPS risk compare to wildfire risk. When we compare those two, we find certainly much higher wildfire risk than PSPS risk, really an order of magnitude or 10 times more on average in general wildfire risk over PSPS risk, which isn't surprising considering what happens when a wildfire starts versus certainly not an insignificant risk of what happens when we have to de energize a circuit. If we can go to the next slide, this next slide basically just gives you a sense for, you know, how we operationalize it or how we manage the data behind the models. There really are two components to it. One way we manage the data is by bringing it into our TechnoSilva viewer.
So you've heard about TechnoSilva from a consequence perspective, but they also provide a GIS viewer, which stores their values. What we've done with TechnoSilva is that we also have arranged with them that we can provide them our probability of ignition values, which they upload into their system so that we have access in this GIS viewer. So we can go to decided
Speaker 6
a full demo of it because of time, but essentially, we provide them updates that were probably admitting data. They allow us to visualize that along with their consequence data. And they also give us a lot of granularity around their consequence data. So we have options. One advancements of that we've been able to make using Tecnosilva overreacts is that Tekmosilva gives us a lot of different options, whether we want to see a max value or 90 percentile value or average value at each specific location.
The other advancement, which was a really important advancement over Reacts, was that Reacts provided data in a raster format. With that, but because we are at an asset level with our probability of ignitions, that required a data scientist to basically extrapolate or interpolate the value from the raster to the nearest asset. TechnoSilva overcomes that by providing us the values at the specific asset level. If you go to the next slide, I'll just talk a little bit about the future improvements. So prioritization of mitigations.
One of the things that we haven't gotten to yet, one reason is because it hasn't been as critical yet, but it is going to start becoming more critical as we move forward, is that as I explained, we have each asset has a curve, a risk curve, so we know for all of the independent assets what their risk profile looks like. Covered conductor, in general, provides the most risk buy down in our service territory. But because it's a curve, you get a lot higher risk buy down in the highest risk segments, and then they start to taper off very much like the curve you saw on Paul's presentation. So there becomes a point where various assets' risk curves start to cross. So, you get to a point where maybe installing a spacer between a line becomes a better mitigation option than going to the next segment of covered conductor.
So, one of the things we're working on moving forward is how do we know when these risk inflection points take place? So, when do we know where, at what points do we know it's better to now move to a different mitigation versus a covered conductor? And how do you combine various mitigations at a location to get the most risk buy down at that location for the cost involved. In general, particularly in the highest risk areas, there aren't we haven't found any combinations of mitigations that will actually be will buy us more buy down more risk than covered conductor, particularly because covered conductor has the unique ability to mitigate both equipment failure risk, along with the contact with foreign object. So, other mitigations that we have, like spacers, don't take don't mitigate that contact with foreign object risk.
So particularly in areas where the consequences are very high and chance of an equipment failure is also very high, doing these alternative mitigations don't really get you anywhere near the amount of risk you get with the covered conductor. The other thing is that's on our plate for this year is really understanding how covered conductor and other mitigations will impact our PSPS thresholds. That's something that we've been studying for a while, and you'll hear more about that, I think, tomorrow during the PSPS panel. But understanding exactly how comfortable we can be raising PSPS wind thresholds as these mitigations are put in place is something that we're studying very deeply right now. And then, as I mentioned, the sequential evaluation of mitigations, how do various combinations of mitigations help?
So for example, what impact do mitigations, particularly around vegetation inspection, what impacts do our mitigations have on those programs? So for example, once we cover a segment, how confident can we be in adjusting our inspection or vegetation programs? Meaning, do we still need enhanced trimming on a circuit that has covered conductor? We want to make sure we understand that very carefully before making changes to the program. So, with that, I think that brings us to the end of our presentation.
I look forward to the questions during the Q and A session.
Speaker 10
right, good morning everyone. My name is Sarah Almujahed. Just checking my audio.
Speaker 10
All right, thank you. So I'm Sarah El Mujahed, I'm the Strategy Manager in SDG and E's Wildfire Mitigation Department. I'm joined by Mason Withers, our risk expert from the Enterprise Risk Management Department. And together we'll walk you through our evolution of risk modelling starting back in 2010, where we implemented the first version of the WORM model, the wildfire risk reduction model, working with TechnoSilva and where we've come today with our WINGS model. So back in 2010, we developed the WORM model to understand the wildfire risk on an asset by asset level.
In 2014, we operationalized this model to give us an understanding of the fire risk potential as weather evolves. So during emergencies, during fire season, we're able to see the dynamic changes in risk modeling using the WORM Ops version of the model. In 2017, we made updates to the WORM model itself to continue to help us inform our grid hardening priorities using different factors. And it's important to note that during this timeframe, there was a lot of changes also in the regulatory world with the risk OIR decision coming out in 2014, our filing of the first ramp in 2016, and in 2018, the first SMAP settlement agreement that established the latest standard for building our risk quantification framework. In 2019, we filed the first ramp using this new risk quantification framework.
And in 2020, we developed wings building on all these milestones that we've come across over the past decade. Next slide please. So at a high level, we thought it would be helpful to just to give you an understanding of our risk informed decision making approach and highlight the three topics we're gonna get into in today's presentation. The first part of the process is really to understand the baseline risk. And that is basically based on using our risk quantification framework to understand the likelihood and consequence of a fire risk and a PSPS risk.
The next step is to identify the mitigation. So, we catalog the mitigations that we have and ones that we are planning to implement as we get ready to put together our wildfire mitigation plan. And then we evaluate the mitigations by calculating risk spend efficiencies that give us an indication of the cost effectiveness of our programs. And the last step here is really dynamic, meaning that it really helps feed into the prior steps. Applies to the baseline risk.
It helps us understand the mitigations better, but really this initiative level prioritization is a refined and deeper dive into specific programs that are so large that we need to have a refined approach to really prioritize the work within. And this is really an example of that is our grid hardening effort where we needed to come up with specific models to prioritize our targeting and which assets we wanna replace first. The key models at the bottom and frameworks that we'll touch on today are three. We're gonna talk about the risk quantification framework. This is really the foundational framework that we use to understand the baseline risk and we use to calculate the risk spend efficiencies.
We'll touch on the wildfire risk reduction model. And this is what we've used historically to prioritize our traditional hardening efforts or what we call bare conductor hardening. And we'll touch on our future model, the WINGS model that we developed in 2020 and how it's gonna help us optimize grid hardening solutions in the future. And with that, I'll turn it over to Mason to walk you through the first two models, and then I'll take over for the last piece to walk you through WINGS. Next slide.
Let's skip two slides. Go ahead, Mason.
Speaker 3
Good morning. My name is Mason Withers. I'm with the enterprise risk management team. We thought we would cover the risk quantification framework a little bit, just so that everybody understands how it's being applied. In the RAMP proceeding and the S map that precedes it, we were asked to create a risk quantification framework using a multi attribute value function.
At SDG and E and so CalGas, this is the one on the left you see at the top, you see attributes, health and safety, reliability, financial and stakeholder impact. We're not gonna dive into all the details of it, but essentially it's a tool that we use to identify how to quantify risks at the enterprise level. So each enterprise has a enterprise risk register and one of those risk registry items is wildfire. And so we basically figure out what's the likelihood and consequence of the wildfire risk on an annual basis, run it through this framework, try to understand the health and safety implications, the impact of stakeholders, how it affects reliability, financial and so forth, and create a single number from that process. Over on the right, you can see our top five risks at SDG and E ranked one through five by the risk score.
So you see wildfire involving SDG equipment is ranked as number one, and then electric infrastructure integrity and so on. The lower is the concept of how likely is the event to happen. SDG and E is using the likelihood of an ignition as its risk event. So lower stands for likelihood of risk event. So our risk event is the ignition and we get about 22 ignitions on as our five year average per year ignition.
The core is the concept over on the left. If an event were to happen, how bad would it be? How much safety, liability, financial implications would it have? And that's the core. So you run those assumptions through the risk quantification framework and you get a core value.
The risk score is the lower times the core. What we're showing this year is because PSPS is such a big impact to the entire wildfire risk, we're showing both the wildfire score and the PSPS score. So on the left of that slash is the wildfire score or the wildfire lower and the wildfire core. And the right side of the slash is the PSPS lower and core. And you can see the total score is 18,000.
So for example, the way you can read it is PSPS. On average, we expect about four PSPS events per year for PSPS. So that's essentially how we calculate the risk scores. And that risk score is important not just to show like how it ranks amongst other risks as a company, but that is the risk score that we say, hey, if we were to do an activity of some sort of mitigation, an asset replacement, how much does it move that risk score? So when we do a new mitigation or proposed mitigation, we recalculate what the law and the core would be if those were in place and we would get a new risk score.
And that's how you develop the RSCs, which we'll get into in just a second. Down the bottom right, you can see a breakdown between the tier two, tier three concepts. If there's any questions on that, we can go into that. For the purposes of saving time, I'd like to move to the next slide please. And here we have an example of how the risk quantification has worked.
Again, I'm not gonna go through every single detail, but essentially what you do is you say, there's a pre mitigation score, and then there's a post mitigation score. So in this case for hotline clamps, that's those connectors between some of the overhead system that they have a certain likelihood of failure. And if we replace those hotline clamps, how many ignitions could be reduced if we were to do that? So in this case, we're saying for $2,000,000 we could reduce 0.008 and we would get twenty five years of benefit for doing so. So that would be, we would apply those assumptions to the post mitigation work.
And we would in the end generate an RSE at the very bottom you see it's 58, which is the change in the risk score multiplied by how many years we get the benefit divided by the cost of doing the project. So we would say the RSE for hotline clamp replacement is a score of 58. I know it's pretty quick run through, but hopefully you'll see in the slides that you'll be able to learn from how we do that. Let's move to the next slide please. Okay, before we, yeah, okay, that's fine.
So Sarah was talking about how we have a current and we have a future. So we wanna make sure everybody doesn't confuse the two, but we're sort of, we got one leg in the past and one leg going forward. So we wanted to show both sides. The worm is sort of our, where we stand right now and we're building upon it. And as the other utilities have mentioned, and it sounds, it's gonna sound very similar.
We essentially look at the likelihood of ignition occurring at each asset. So we look at each pole and poles have conductors attached to them. We're saying, what's the likelihood of a failure which would lead to an ignition on each pole? Then we say, how bad would the consequence be if we were if an ignition were to occur there? And just like, I think it was Paula said, it's simply the likelihood times the consequence.
And that gives you your risk score at each pole location. So in the past, what we would do is we would take that risk score and we would say based upon that risk score, we can evaluate how proposed projects would look if we were to do so. So just like everybody else was saying, we use different inputs to try to figure out what the likelihood we are using Technosilva to come up with a consequence and we marry those two together to figure out where we should do projects. Let's move to the next slide, you'll see an example of how we've been using that a few years. So here's a visual, get a little feedback.
Here's the visual of the worm output. Each pole on the right, the graphic on the right, each dot represents a pole, and it's already had its likelihood and consequences sort of determined or estimated, I should say all models are estimates, estimated. And then we're color coding in such a way that the high risk ones are colored red, lower are yellow. So what we would have done with this in the past is we would give this to some engineers and some project managers, designers to say, hey, what can you do on these areas that could make the likelihood go down? Typically these jobs are not reducing the consequence.
If the fire did start, it would probably burn the same. So it's focusing on the likelihood. And so we would give it to some project managers, they'd figure out an area of lots of red dots and say we could do the following in this area, which was typically hardening and maybe putting some improved infrastructure in there. And so we would go through that process of trying to reduce the risk at the best usage of our dollar based upon the word model. Again, so that's where we are as to right now.
Sarah is going to talk about where we're going in the future. I think that covers my slides. Thank you very much.
Speaker 10
All right, thank you, Mason. Let's jump to the next portion of the presentation to talk about wings. Let's go to the next slide. Okay, so before we dive into wings, we thought it would be helpful to just kind of level set and just understand this concept of different levels of granularity for models. And if you think of the spectrum of granularity on the one side on the left, you might have modeling approaches and analysis happening on an asset level.
And on the far right, you might have analysis happening at an enterprise level. What Mason presented really covers those two ends of the spectrum where he talked about the enterprise risk assessments using our risk quantification framework. That's what goes into our ramp. On the other end is the asset level analysis that we've been doing through WORM to quantify grid hardening projects and prioritize those. Now, with the heightened focus on PSPS, we found a need to come up with a layer of granularity between those two opposite ends.
And we defined this layer as the segment level of granularity. And a segment is defined as a collection of spans and structures between two isolation points or two SCADA devices that we operate during PSPS. We had to look at the system as a grouping of these assets between isolation points because when we look at PSPS and when we de energize, we don't de energize an asset. We don't de energize a pole or a wire. We isolate portions of a circuit where we see high wind conditions, risky situations.
And so, we had to understand the PSPS risk from a segment standpoint and be able to compare it to the fire risk at a segment standpoint. Now, one point I'll make here is just because we created this layer of granularity doesn't mean we're gonna stop doing things on the asset level or the system level. All of these layers feed into each other. We actually use the asset level information to inform how we come up with the assessment for the segment levels. And that's where we came up with WINGS in 2022 specifically do this type of analysis.
The one thing I'll mention about WINGS is that it's a modular model, meaning that it can take inputs from various sources and models, and we can continue to improve those modeling approaches outside of it and feed into it. So, it's really flexible. It's not set in stone, and it can take what we've done from the asset level, and it can take what we're going to do with independent studies as we continue to evolve our modeling capabilities. Next slide please. Okay, so with wings, what it really leads to is a shift in strategies for how we want to mitigate our segments.
Because as I mentioned, PSPS is a risk that we wanted to account for with the development of wings. And as I mentioned previously, looking at things from an asset standpoint is good, but in the context of PSPS doesn't help us a whole lot. And that's why we created WINGS to really shift from an asset level approach to a comprehensive segment level approach. And what we're showing on the left here is a visual to give you a bit more understanding of what that really means. So the first graphic on the left is a geospatial mapping for one segment prior to using wings.
So you'll see this one segment has a mix of wood poles and steel poles. And so, in extreme situations when the fire risk is heightened, we on the conservative side. We know this segment hasn't been completely hardened. So we may not feel comfortable enough to operate it under high weather conditions. On the right hand side, we're transitioning towards this whole segment solution approach so that we can account for things like objects flying into the line, really mitigating the entire segment based on the situation of that segment.
Again, as I mentioned, our asset level analysis really feeds into this and informs it. It doesn't take segment level analysis doesn't get rid of that granularity. And that's what I'm highlighting on the right where I'm showing the worm simulated asset analysis feeding into WINGS. So, every poll on this chart essentially has a WORM risk score and we use the components of that risk score to really aggregate the assessment at a segment level. Now, facts about WINGS, and this will continue to evolve, but where it stands today is we have over 600 distribution segments being assessed in WINGS, comprised of 90,000 poles across the HFTD and a little bit outside the HFTD with around 3,600 miles of distribution overhead.
And the average length of a segment is around six miles. The benefits of WINGS, as I mentioned, is it helps us reduce the PSPS impact by identifying whole segment solutions. It builds on prior models and integrates their outputs, and it utilizes the same MAVF or the risk quantification framework that Mason walked us through that we used for RAMP and the WMP. Next slide, please. So what WINGS helps us answer is really three key questions.
It helps us understand the current risk level today, the baseline risk for every segment. It helps us do this alternatives analysis using the risk spend efficiency to calculate risk scores for different types of mitigations. And as of right now in 2020, we used it to look at what a segment how the benefit could look like if we underground a segment, if we applied cover conductor, if we gave customers generators, or if we continued our traditional bare conductor hardening. We used each alternative risk spend efficiency score, and we're using that to inform our approaches in the future.
Speaker 10
And then the last question that it helps us answer is, really, it helps us to play out different scenarios in a top down fashion. Meaning, we might have SMEs wanna play out a different scenario, and then we can go back to the model and see what that scenario could look like. Scenarios could be significant hardening in certain areas of the territory, limited hardening, focusing on one type of mitigation, giving customers generators, and we're able to go back into the model and actually quantify the cost and the benefit of these different portfolios and see where we get the most the biggest bang for buck essentially. Next slide, please. All right, so inputs and outputs of the model.
So, as many of my colleagues today in the prior sessions talked about, risk is comprised of likelihood and consequence. And WINGS helps us understand both the fire risk and the PSPS risk. Each of those, we have different data points to quantify the likelihood and the consequence. For the wildfire risk, we're really tailoring the risk assessment to the unique conditions and the characteristics of the segment. So, we're looking at historic ignitions, we're looking at wind speeds because we have weather stations mapped to SCADA devices or segments in our situation.
We have an understanding of the length of the tree strike. So, out of the entire length of the segment, how much of it is really at more risk of hitting a tree or being struck by a tree. We know the hardening status because we've done a lot of hardening work that really lowers the risk, but some segments are not that significantly hardened because we've been focusing on specific assets at risk. And the three last bullet points under the likelihood for wildfire are three new items that we added towards the end of last year, which is understanding the vegetation density on that segment, taking into account the circuit health index, which gives us a little bit more information about the specific assets in that segment, as well as the conductor age. And then from a consequence standpoint, as I mentioned, we're building on what we've done in the WORM model.
So, we're using the WORM conditional impacts that tell us at every poll what is the potential consequence of a fire, where to start there. So we took every poll on a segment and we aggregated it for this assessment to come up with the overall consequence of a fire risk if something were to happen on that segment. From a PSPS standpoint, we're looking at the annual red flag warning data. We're looking at historical wind speed patterns. For every segment we have a probability of shutoff that is based on the unique conditions, as I mentioned of each segment, whether it's the wind or the vegetation, the level of hardening.
And then one of the things that we added most recently is this concept of circuit connectivity, which was really complicated at first. When in our first iteration of the wings model, we were looking at segments in isolation of each other. And that resulted in really suboptimal mitigations when we looked at the outcomes and we looked at the segment in the context of a circuit, we would have solutions that were pointing us, for example, to underground the most downstream part of a circuit without mitigating the upstream risk. Well, that's great for that segment, but if a shutoff happens upstream, the customers don't get the benefit of that underground. So we've updated this so that the model now actually will run a combination of scenarios and mitigations for every segment in a circuit and come up with the right mix of solutions that maximizes the benefit for the customers and from a fire risk reduction standpoint.
For the consequence of PSPS, it's pretty straightforward to some extent. We're looking at the number of customers that could be impacted by a shutoff in any given segment. And we're looking at different customer types. This is something that we're continuing to evolve. But our latest update on it is we've dissected the types of customers into three categories.
The first category is critical customers. This is our urgent and essential customers, hospitals, fire stations, police stations, communication towers. The second category is medical baseline. It's pretty self explanatory. And the third category is a catchall for any customers that don't fall in the first two categories.
And we call that the non critical customers category. But that said, we are continuing to refine this and we'll have more granularity with regard to the customer types. We're looking at divvying up the residential, commercial, industrial. We're looking at better understanding the AFN population impact. So, this is a continuing work in progress that we're going to update this year.
The outputs of the model, this is a bit of an eye chart, but just to give you an idea of what comes out of this model. Using the WINGS model, we can look at the ranking of segments based on risk. We can look at the ranking based on risk spend efficiency that'll tell us where we can start the work to get the most benefit out of our projects. We can look at the RSE analysis, as I mentioned earlier, to understand the different alternatives and the cost benefit of those and form our decision making that way. And we can also look at the portfolio analysis, which really aggregates the outcomes from every segment, adds it all up and gives us these curves that really tell us and help us understand how good is the risk reduction on a per dollar basis.
You'll see examples of portfolios that have diminishing returns, but it's just being able to visualize it this way that really helps guide our strategies in the future. Next slide please. So some quick applications of the WINGS model in 2020. We were able to pivot some of our mitigation strategies in the 2020 WMP, meaning the timeframe of the 2020 WMP 2020 through 2022. But really pivoting in flight work is not that easy.
It takes about twelve to eighteen months to go through a project from scoping to design and construction. And so, really a lot of the work that we're doing with wings is not going to be implemented until later in 2022 to some extent, but mostly in 2023 and beyond. That said, we were able to pivot some of the work depending on where it was in the design phase. So, we're showing an example of an output from wings for one circuit on the left side, that's circuit four forty eight, and all the segments within that circuit. And we looked at the different alternatives.
So, cover conductor, that's the CC risk spend efficiency. We looked at undergrounding, and that's the risk spend efficiency for undergrounding. We looked at continuing traditional hardening, and there's the risk spend efficiency for that. And then the recommended mitigation is where we landed and the reason for deviating from the highest RSE. Generally speaking, the higher RSE, the better is the project.
In some cases, as you'll see here, for some segments, we weren't able to pivot, but we wanted to validate that we were still doing good projects in terms of their risk scores. And that was really validated by the risk spend efficiencies that you see. However, for some, for about three of these segments on this circuit, we were able to pivot because we weren't that far in the design phase. And we're actually now designing them for the implementation of covered conductor based on being the highest risk spend efficiency for these segments. On the right hand side, we also did this validation just to kind of, know, WINGS is an improved model.
It has a lot more elements than what we used to look at in WORM, but it was really comforting to see that the highest risk circuits are the same ones that we had been targeting. But as I mentioned with WINGS, we're going to be gaining more benefit because we're looking at the combination of PSPS and fire risk, and we're using more up to date information. Next slide please. So just some closing thoughts for today. Based on the discussion today and our timeline, we have a history of using risk models to inform our mitigation efforts.
We think risk modeling should continue to evolve based on emerging challenges, and there needs to be a flexibility to determine the appropriate models to use for various fun trips. It's really not a one size fits all. SME input remains crucial to making the final decision. There are things such as feasibility analysis, land permitting and other constraints that may lead us to select different mitigations from what the model may point to. And as I mentioned, it takes time to transition to new frameworks and shift operations and decision making to new standards.
Our next steps is we're going to continue to update WINGS this year. We're going to continue to refresh our assessment and continue to inform the grid hardening in 2023 and beyond using this WINGS model. We're going to use WINGS to explore the potential application of evaluating vegetation management on a segment by segment basis. And we'll see where that takes us. And there will be additional enhancements that we're going to identify this year and develop a roadmap to continue to improve it.
And as I mentioned, WINGS is really going to primarily inform the scope of work in 2023 and beyond. And with that, I will end my presentation. Thank you.
Speaker 4
Thank you for all the presenters. So now we are going to move into a break. So we are going to resume the question and answer segment of this year starting at 11:15. So we will all be back here at 11:15. I will also post that in chat, and then that's when we'll resume the question and answers.
Thank you to all the presenters again. All right, everyone. My clock is reading 11:15. So let's go ahead and get started with the question and answers section of the risk assessment mapping and resource allocation methodology
Speaker 4
All right, so just a really quick note, as we start the question and answer, there's gonna be a lot of active discussions. So participants and people in the audience, be aware of your mute button, just so we don't have a lot of feedback going into the conversations. Joining us in the question and answer section, we have a few stakeholder parties here. We have representatives from TURN, which is the Utility Reform Network. We also have representatives from Mercy Grade Road Alliance.
We also have representatives from Ring Power Institute, and we also have Mr. Abrams in the panelists as well. So stakeholder participants, what we'll do is before you ask your question, go ahead and say your name and what party you're associated with, and then will have the questions started. And then joining me as well is going to be Melissa Simpson. She's going to be moderating the chat and let us know the questions that have been coming through chat.
Cool, so let's get started here. I'll go ahead with the first question and then I will go ahead and let the panelists know who is next. The first question is coming from WSD, and this is just to set the stage for this next question and answer section. We've all seen an overview of the risk models. We wanted to just have the utilities highlight the specific changes that have been made since the 2020 WMP and how those updates will results in improved ability to model ignition risk and wildfire spread.
So it's a two part question, the specific changes and also how those changes will improve in the ability to model ignition risk and wildfire spread. And then the order that will go is PG and E, SEE, and SDG and E.
Speaker 5
Hi, thank you for that question. I think that really gets to the crux of why we're meeting today. From PG and E's perspective, initial risk models were built in late twenty eighteen-twenty nineteen and we have changed absolutely everything in those models since the during the 2020 period to be in production for 2021. In my presentation I discussed it and we I think we did quite a good job of discussing that in our wildfire mitigation plan. It's in section four, common to the other utilities as well.
The main things that we've done is that we changed out the initial probability component and we've utilised this maximum entropy approach using a machine learning type analysis. A lot more data went into it this year. We used the twenty fifteen to twenty eighteen CPUC ignition data set the learning module and then to the testing component of it we use the twenty nineteen ignitions. So we're predicting using one set and then testing against reality using the other. What does that allow us to do?
We've got a bigger covariant pool, our focus is on vegetation and on conductor for the two main failure modes. I did see a question on that and I think we'll probably get to that one a little bit later. What this modelling approach allows us to do is make a better accurate prediction of where and when such an ignition may occur and then allow us to pull in the various covariates together and have a look at those in a mixed setting. The second half of our model that changed was that movement from the REACTS fire propagation and consequence module to the Technosilva fire propagation and consequence module, and there's a lot of changes in that itself. Technosilva sits on a more updated platform than the REACTS model did.
I'm not saying it's better, I'm just saying that it's different. REACTS was constructed using a 2014 LANDFIRE data set, which is the ground fuels from 2014, but those images are actually from like 2011 or 2012, so slightly older. Tecnosilva sits upon a 2016 LANDFIRE data set and has been updated to include all disturbances to 2020. So when I say disturbances I mean fires, so updates to the fire module. The building structure and the population data inside Tecnosilva is an updated layer to React.
Tecnosilva is using a 2018 Microsoft update. React engineering used the 2010 census data, so again we're seeing differences there. So when we put these two together we're seeing a lot of changes between our model. The third part was PG and E endeavoured in 2019 to use its first version of its egress model. That was included in its 2019 and 2020 wildfire mitigation plans to prioritise.
It is not something we're using for the 2021. It's something that we'd like to develop in partnership with the WSD and the other utilities as we go forward in 2021. Now coming back to key points, we changed out the probability model from the components of conductor and vegetation and we changed out the consequence model for an updated data set and we also changed the algorithm. So everything in PG and E's model has changed. We are seeing differences.
Those two differences are mostly attributed to the fuels layers with regards to what we call surface fuels, ladder fuels and then crown fuels. Crown fuels are the tops of the trees which have the highest heat rate, so we saw a lot of progression of change attributed to that. What it's allowing us to do is it's allowing us to focus on areas where we would expect an ignition to occur based on the attributes of the environment and the asset, as well as forecasting where that fire would go based on surface fuels moving to ladder fuels moving to crown fuels.
Speaker 7
So for SCE, I won't try and repeat everything that Paul said, some of the changes are very similar. Our major changes were moving from REACTS to Tecnosilva, and it has the same benefits that Paul mentioned. We were also able to a big change that we made was in our 2020 WMP, we did all of our RSE calculations on a system level, so basically there was one RSE calculation for each mitigation. This year we were able to get all the way down to a much more granular set of RSEs as well, and what that allowed us to do from an operational perspective is to sort of reorder the risk methodology to see where each segment's tied up, and what that does is gives us a better way of targeting our mitigation specific segments that have the highest risk based on a new methodology. And the last thing is, as I mentioned in my presentation, is we've created the PSPS risk stack, which is brand new from 2020, didn't exist in the 2020 wild mitigation plan, and gives us more visibility to where PSPS risk exists on our system in terms of impacts to customers.
And one thing I neglected to mention during my presentation is as part of that risk stack, we've created an access and functional needs and a critical infrastructure adder that we use in that formula, which is which enables us to identify certain circuits that have higher impacts to those populations, and it elevates those circuits on the PSPS risk stack. So there's several we made several major changes that have allowed us to look at risk differently.
Speaker 10
And for SDG and E, I'll highlight three key changes that took place in 2020. The first one is the movement from an asset level assessment to a segment level assessment, as I mentioned earlier, and a big part of that is taking into account the PSPS risk so that we can hit two birds with one stone for lack of a better explanation, but it helps us really optimise our mitigations to reduce both risks. The second thing is we improve the input to the likelihood. So, where WORM was really focused on likelihood of assets failing, WINGS really takes into account a lot more data points to look at the likelihood of ignitions from vegetation, the wind conditions. And one of the biggest things we also improved in 2020 is we went back and conducted efficacy studies on our grid hardening efforts to date, and actually incorporated those estimates in our analysis and wins to determine the effectiveness or better estimate the effectiveness of mitigations.
And then the last thing is really using an updated multi attribute value framework. That's the latest implementation of the risk quantification framework that Mason described. And in 2020, the work will continue. We're continuing to update it. We're continuing to analyze the risk.
And as new information comes in, we'll be able to better refine that model.
Speaker 4
Great, thank you so much. And then I did want to also say that earlier when I was going through the stakeholders, I did forget to include, we also have representatives from CAWDA advocates as well. So let's go through some of the other questions here. So the Utility Reform Network, TURN, go ahead and unmute, introduce yourselves, and then ask the question. Thank you.
Speaker 11
Great. This is Katie Morsoney from TURN. Before I get started, I'd like to raise the same point of order that Joseph did. I have a variety of questions. Should I run through all of them, or should I go one by one?
Speaker 4
Yeah, thank you for that question. So this will go for all the stakeholders. Let's keep the questions to one at a time, and then we'll go around in circle. Totally fine if your one question has multi parts, but let's try to just tackle one at a time, and then we'll make our way around.
Speaker 11
I have a question. I'm going to start with PG and E. And that is, this is a fairly simple question. Is my understanding correct that you're using a different process for risk assessment than we recently have been discussing in your RAMP filing?
Speaker 5
The simple answer to that is no. For the 2020 RAMP filing, PG and E uses an enterprise level risk model which looks at risk at the portfolio level, so at the highest level in the company, and we developed the multi attribute value framework to be used at that level. For wildfire risk, we're following that process using the core and the lower, and we have the risk equals probability times consequence. So it's a very similar process. What we're doing at the wildfire model is we can take this down to the circuit segment level.
And then on the output side we can use the outputs in either natural units that come from technosilva or we can convert those back to our MABF, which is aligned with our enterprise risk model. Did that answer your question, Katie?
Speaker 11
So okay. Sorry. I'm I'm thinking and talking at the same time, which is always dangerous. So you aren't using the same model because in the RAMP, you were looking at an enterprise level and here in wildfire you're looking at a more granular level?
Speaker 5
That's correct. At the ramp level it's enterprise. So we're looking at wildfire as the risk. There's a number of drivers feeding a risk event with a number of potential outcomes. And then we're summing all of those together to get our total MAVF score and that is for the risk.
What we're doing in our wildfire mitigation plan using our wildfire risk model is to take that risk event and break it down to specific circuit segment levels. And let me define what a circuit segment is. A circuit segment is a component or a part, a subset of a circuit. They range from about, you know, 0.02 miles to about 12 miles long in length. The average is about six miles long and we can tend to be able to control these bisectionalisation devices.
So when we're looking at wildfire risk in the wildfire mitigation plan, we're looking at a circuit segment level, not an enterprise level.
Speaker 11
Okay, thank you. I have other questions, so I'll wait for my turn again.
Speaker 9
Hi, this is Joseph Mitchell from Massey Grade Road Alliance. This question is for all the utilities, but
Speaker 9
to start with framing it with PG and E because I heard Paul say is that they do have sort of a time dependency in their probability of ignition function. What I've read in the WMP is that their probability of ignition function was location dependent, and not so much specifically time dependent. So to be specific, do your probability of ignition models take into account the contingent weather conditions at the time of the risk event? Or do they, for instance, take into account the peak wind or peak weather conditions at that location? Or do they use an average weather condition?
Because what I had read in the PG and E's WMP is that they're using averages, and that doesn't seem right to me because wildfires don't occur under average weather conditions, they occur under extreme and peak weather conditions. So how do you take into account, with probability of ignition, these the weather contingencies on a specific event by event basis, on a peak overall peak for a location, or as an average?
Speaker 5
Okay, thank you for that question, Joe. There's a lot to unpack there, so let me try and unwind some of that. Our modeling is based on asset data, environmental data and weather data, all of which has a temporal component to it. There's age of the asset, there's age of particular upgrades or changes out, with regard to environmental information it changes across the course of the year so we're using different time periods when we're assessing that and for weather conditions as you correctly say weather changes and wildfires tend to occur during adverse or high fire threat wind type periods. So that is in the model.
With regards to the prediction themselves, we're using the training set of data which looks at historical CPUC reportable ignitions and they have a temporal flavour to them in that we can look at the distribution of those ignitions over the course of the year. The majority of them happen in what we call fire season which tends to be between the period of late May, early June through the November. That doesn't really don't happen outside of that, that tends to be when the majority of those ignitions occur. So when we're looking at these ignitions and we're making these predictions, there is a temporal component to it, but there's also that locational specific component as well based on the environmental conditions at that particular place. So there's a lot of things going on, the particular algorithm that we're doing doesn't rank them in rank order but more of a permutation type order.
In other words, if we get a small bore conductor with splices on it in an area with high grasses, high winds that are dried out, there is a more likelihood of an ignition occurring in that particular geographical location at a particular point in time.
Speaker 9
Yes, I looked at your the variables that were used, and I saw average wind condition. I did not see peak wind condition in the list.
Speaker 9
I can take that answer offline, or, you know, we can handle it separately if you don't know off the top of your head.
Speaker 5
I don't really off the top of my head, but I can get back to you if I'm just looking at the list of the topic, the top covariates that were in our conductor model. Let me get back to you on that one.
Speaker 6
I can go next if you want. This is Joe Boyceweta, Southern California Edison. So let me break it into two parts, the consequence and the probability of ignition. We mentioned earlier when we went to Tecmo Silva, it gave a few more options for consequence, specifically whether we want to use the max, the 90 percentile, or the average. So in general, we have the ability to look at these two components in different ways, and we're going to usually choose what makes the most sense for the particular problem.
So for example, for planning mitigation work that may take a year or a year and a half before we can actually complete the work, we're likely to use the max consequence. That ensures that we're going after the areas with the highest consequence in general. On the probability of ignition side, we looked at it a few different ways. First, we did correlation analysis at the circuit level with outages and wind. So we originally thinking that, you know, do we really want to build our model off of, you know, to avoid high wind situations?
The challenge was that the correlation between actual wind speed and when faults occur is fairly loose. It's not a very strong correlation, particularly when you start getting down to the asset level, circuit level. When you start bringing it up into zones or districts or regions, you start seeing a little bit more correlation between outages and wind speeds, but the correlation isn't super strong. Plus, if we started limiting the development of our models to specific weather scenarios, like only data that occurs on high wind, high FPI days, we start making this tradeoff between model accuracy that you get from having lots of data and less accuracy because you're using smaller and smaller sets of data. Also, because, again, we were building, mostly we're building these models for the purposes of prioritizing mitigations, which are going to take some time to advance, and also because, you know, we've now seen that you can have very large fires even on relatively low wind days from these fuel driven fires.
The idea that you only get big fires when the wind is very strong, last year I think was proven not to be correct. So what we do is we build our, we want to get at the asset level, and we want the model to be as accurate as possible, and so that which means we have to use as much data as possible, we built our models, our probability of ignition models using, like I said earlier, outages that could lead to a spark, and using all outages. So in practice, what that actually ends up meaning is that our probabilities of ignition aren't only, aren't specific to just the highest wind days. They basically are, they represent what is the current probability of ignition for each asset in our system. And then the way we do a time component to that is through some calibration, which if you were interested in the details, a data scientist could tell you how that works.
But we calibrate it to get a sense for, based on our system and our probabilities, how many ignitions would we expect in the next year, for example. Did that get to your question, Joe?
Speaker 3
Joe. This is Jason. Good to hear from you. I'll try to make this brief. I think we're pretty similar to what you've heard from the others.
Because this is a long term planning model, which is different than what we do in operational modes. We're looking at the max and we're looking at the max wind speed and the max consequence right now. We know a perfect world we want to extend out like an 80 seven-sixty view of the year and have the lesser events included as well. But right now we're looking at the max. So the roadmap is to have a bigger view, excuse me.
But right now we are looking at max.
Speaker 9
Okay. So that's not on an event by event level, that's sort of like an annual MAX or something like that?
Speaker 3
Correct. For the time being. In 2021, it may no longer be that way, but that's how it is at this moment.
Speaker 4
Great. Thank you, Joseph. Yeah, let's move over to Green Power Institute.
Speaker 13
Hi, this is O'Hara with GreenPower Institute. My question is for PG and E. So as has been mentioned a couple of times already, the machine learning models that train on relatively large data sets are more accurate, and so SCE is using the outage data set to increase the number of events in their data training input. In contrast, Recy, PG and E is using ignitions, which actually has one to two orders of magnitude less data than the outage event data, and then two to three orders of magnitude fewer events than the total risk event data set. So my question is, why is PG and E using the ignition data set that only has in the tens to thousands of events inputs versus richer event risk data sets?
Speaker 5
Thank you. Hi Zoe, thank you for that question. That's a good one and thanks for bringing it up. In the 2019 and 2020 models that PG and E presented, we also used outages as the dependent variable in our regression. The independent variable variables being the asset attributes, etcetera, etcetera.
And we were using that as outages. What we find, and again this is related to the data as you correctly say, when we analyse and examine the particular outage data, the outage data tends to be collected by the nearest piece of recording equipment as opposed to where the particular risk event occurred. So when we look at ignitions, we tend to have the lat long of those ignitions because we've had a troubleman or utility personnel in that particular area. So we can be more specific to our actual area of where the risk event, the ignition occurred. PG and E did make that movement.
We went from outages to ignitions. It's predominantly around identification of the specific location of where it occurred. Our particular granularity for this latest model, the 2021 wildfire distribution risk model, is at 100 metre by 100 metre resolution. If we went back to outages, we would be moving that to a lesser pixel granularity, and it would be at the piece of equipment that's picking up that outage.
Speaker 12
This is Henry Burton with the Public Advocates Office. I just want to follow-up actually on Joe Mitchell's question about weather scenarios and the risk modeling. So in a previous workshop, Edison said they have 41 weather scenarios and they picked a scenario with the maximum consequence. And so this is a question for Edison and also SDG and E. When you say you pick the maximum consequence, do you mean the scenario that has the greatest consequences across your entire service territory?
Or do you pick the maximum, the scenario that yields the greatest consequences at each asset level? Do you understand what I'm asking there?
Speaker 7
I think so. This is Robert, and Joe can chime in and correct me if I get this wrong, but my understanding is our the 41 weather scenarios are actually, you know, put into an algorithm that, like a Monte Carlo type simulation, and so it's not that we just pick the max, but we pick the max forty one days for a particular location and then run that through the simulation, right, to generate the score, so it'sI don't believe that it is picking just one number, but Joe, if you have a different response, could correct me if I got that wrong.
Speaker 6
Yeah, would just I'd add to it. So if you remember, we mentioned that TechnoSilva is providing us consequence values at the asset location. So that means a simulation is run at each asset location, and we take the max consequentity at each location.
Speaker 12
Okay, so that means you could have different weather scenarios being employed for different assets?
Speaker 3
Yeah, so currently what we're using is the max at the location. So there will be different wind patterns and climate patterns at each pole. And we're taking the max at the pole. But as I mentioned to Joe, though, that's changing and we're going to be taking into account more. Right now we're taking the max at each location, not a system wide max.
Speaker 12
Got it. Okay. And PG and E, do you have an answer to that question? It wasn't clear to me that you were using the same type of approach.
Speaker 5
We're using a similar approach, Henry. So again, on our probability model, and this is the question that the answer that I owe Joe back, I believe we use both average and max for the wind speeds and the inputs to generate the probability. On the output we all use the Tecnosilva application. Tecnosilva can give us those answers either at the maximum consequence the P90 or also the mean. It's my understanding PGD is using the P90 when we look at that, so a slightly different approach but again it's just a flavour of which we can take.
We can see all of the results of what we're using in our modelling and again Mason actually said this really nicely earlier on is all of our wildfire models are for system planning and mitigation type activities. These are not real time operational models. The real time operational models that are being used for PSPS type activities are looking at the actual weather versus historical weather patterns. I think that was worthy of noting and thank you, Mason, for making that clear.
Speaker 14
Thanks very much. Great to see progress in the wildfire mitigation plans. I have a broad question before I get into further questions around methodology. The broader question that I have is I sort of see a big gaping hole in these plans, and I'm hoping that folks can either fill the hole or identify that it's not a hole. So specifically sort of tying the assessments to our catastrophic wildfires.
So, you know, as an example, sort of an analogy, you know, if the Airbus went down with Boeing and Boeing went to the FAA to explain why these catastrophic wildfires occurred and built into their risk modeling and the risk assessment improvements based on that, you know, they would identify, you know, perhaps where the fan that was at fault was incorporated to within their assessment. Or the air inlet, or the electrical coupler, or whatever those components are within that airplane that failed, that caused that accident, they will report how they incorporated that into their methodology and assessment. I have not seen in any of the presentations thus far that specifics about, here's where we incorporated the learnings, as an example, from the Kincade fire. This jumper failed. This is where infrastructure needed to be de energized and wasn't, and it is represented in our wildfire mitigation plan here, here and here.
And so, I guess that's sort of the big gaping hole that I see, is tying recent catastrophic wildfires very specifically to how you've addressed them. So I guess my question would sort of be two part. One would be, you know, can you please, I guess, state to me that you don't see that as a big gaping hole, in which case I understand that's your position. Or if you filled that hole, if you can be more specific and sort of go through the catastrophic wildfires that have occurred in recent years within your territory, and explain where specifically that has been incorporated into your wildfire mitigation plan. Thank you.
Speaker 3
This is Mason. I'll take that one first, if that's okay. First of all, Mr. Abrams, nice to see you again. I would say that we are trying to do exactly what you're talking about and maybe we need to communicate it a little more clearly.
SDG and E and I believe the other utilities have done a very good job at researching what the circumstances were that led to large fires and try to understand, was it an asset issue? Was it an outside contact issue? What could have prevented that fire? What were the circumstances environmentally, the vegetation, the engineering conditions? What were the inspection records on everything?
We've been studying wildfire pretty intensely since 2003, when we had some local fires that were not associated SDG and E, but then, I mean, literally for seventeen years, we've been tracking ignitions and trying to understand the conditions that lead to those catastrophic ones that you're talking about. And we've done research, we have a fire potential index, and that was created for the specific purpose. I think what you're talking about, which is in extreme conditions, extreme conditions are probably over 90% of where wildfire risk is. There's an elevated piece, which is a little bit of fire risk and the normal conditions, there's not much wildfire risk to speak of. So we've been doing that kind of analysis for quite a while.
The tier system of tier two, tier three is also driving towards that of doing the work on these focused areas where if the fire started in those locations, they would be larger than if they happen in other situations. So we're trying to look at it from an environmental perspective, from a geographical perspective, and then we've done ignition management to figure out what are causing the outages. So we look at, is there a piece of equipment or was it outside contact? And we've gone through that for years now. And every time an ignition comes up, we try to review, is there something more we could be doing about that?
So for example, back in the day, we had a lot of capacitor failures that were causing ignitions. We tackled those, we replaced all old capacitors in our back country because we didn't want those to be a fire source. We go through asset by asset trying to figure out what circumstances would a catastrophic fire occur and what can we do to mitigate those? PSPF is sort of the final straw along that where we can say, look, with this given equipment, with these wind conditions, we think there is a chance that an ignition could occur. And so PSPS is the last resort, but all that is to prevent catastrophic fire.
I hope I've made my case.
Speaker 14
I appreciate that. And just to clarify, I think part of this just has to do with the audience and the consumers of the wildfire mitigation plan. And so I think that part of the folks who are the consumers of the wildfire mitigation plan are our rate payers, our residents who lost their home in 2017 due to a catastrophic wildfire, who evacuated from their home in 2018, who again had a close call in 2019 and so on. So, you know, part of what those residents I imagine would need to see and the commission see in a wildfire mitigation plan, that I can breathe easier, if I live in these areas is to say, okay, great. PG and E had the Kincade fire.
They identified these three things. They addressed it here, here and here in their wildfire mitigation plan. Now I can breathe easy, because I know they've incorporated that into exactly where they need to do it so that we're safer moving forward. So it's that level, again, for that audience that really needs that. And I understand the high level presentation and walking through the methodology, but unless we get to the specific level, particularly around what caused these catastrophic wildfires, I think we miss the mark with who our audience is.
And we miss the mark in terms of really being more proactive. And if we wait on the CAL FIRE reports, as an example, that still hasn't come out yet from Kincaid, then it again is not really providing assurances that we've addressed those things. And that's the, again, this hole that I see that really needs to be addressed. So if we can, I guess be specific about that, that would be fantastic? Thank you.
Speaker 5
Will, I'd like to take a shot at answering your question. Because I think a lot of that was put towards PG and E and I appreciate those questions. With regards to our wildfire mitigation plan for 2021, it is very distribution voltage class focused as I'm sure you've observed. The thing about distribution voltage ignitions is there are a lot more of them. For PG and E, off the top of my head, it's about 90% per year, so it's about 90 over 100 ignitions per year.
And from my time at the other utilities it's probably a very similar ratio. So we have a lot more data on the distribution side and that's the first order of business that we've gone after. As part of our wildfire mitigation plan and our model build there is a number of failure modes effects and analysis FMEA that goes on and gets done so that we're looking at components that we need to mitigate. For example, and again I speak about this only generally, that we've seen failures related to particular types of connectors or clamps or what have you. And then we analyse those and then we look at taking them out of service whether it be expulsion fuses, that when they go out they've got the reaction that causes the fuse to expel, there's spark and there's heat and they can be a source of ignition as well as protecting the fault.
So we're then now not deploying those in certain areas. So that's our learning. We tend to be writing that into the mitigation type discussion. Also, as you said, with regards to the Kincade fire, and I remember reading about this before I became a PGD employee, the particular piece of equipment that was suspected of being part of this was a jumper cable. And that particular piece of analysis is not discussed in our wildfire mitigation plan because our transmission distribution, sorry our transmission wildfire risk model is currently in development.
What is discussed is what we call the Oh model or the OA model, it's the transmission operability assessment model. You can find a little bit more about that in I think it's section 4.3 where we discuss PSPS de energizations of the transmission level. And again, just to hit on something Mason said, what we're doing is we're looking at real life system conditions, what's on those assets, what's been identified as a potential problem on the operability assessment side of the metric and then on the, you know, the x axis we're looking at fpi fire potential index. If something were to fail, what is the outcome of that? That is in our real time decision making for making de energisation decisions around a transmission circuit.
With regards to our planning models for transmission wildfire risk assessment, we're still building that process in there. So we've got a ways to go on that, as I acknowledged earlier on, it's one of the things we're working on in 2021. Unfortunately, we are struggling with data on that one. It's something that all the utilities could potentially work together and pool their data. We all have similar types of equipment and we can use that larger data set.
As I said, for PG and E, for transmission voltage, it's about nine events per year and subject to check with my SCE colleagues, if I recall, it's about four events per year. So it's a very small amount to do that analysis on. But we're making progress.
Speaker 6
Yeah, and I'll add for SEE. Similar to what Mason described, we have a process to do root cause analysis on every ignition, whether they're CPC reportable or not reportable ignitions. So each time an ignition takes place, we have a process where we collect information on the circuit, a circuit's history, reliability, inspection, maintenance has been done on the circuit. And we also send engineers out into the field to actually look at the equipment that failed or try to figure out what caused it from an engineering perspective. What they gather from that gets fed back to us in the modeling area to enhance models.
So, for example, about a year ago, they started to understand the impact of wind on the lines and how wind and the direction that the wind was hitting the lines could influence failures in multiple different ways. So one of the things we created for our wildfire risk model was the cumulative wind force. We went back to hourly data over the last ten years, and then using basically your basic physics wind force equations, calculated the cumulative wind that every piece of equipment is experienced based on not just the speed of the wind, but the direction that the wind was hitting the lines. We found that was that improved the model accuracy by about 15%. So that's something that we brought in from an engineering analysis.
But on top of that, sometimes the engineers find things that you can't pick up in a model just yet. So for example, when you read this the WMP this year, you'll see there's two activities that we put in there. I'll hit on one of them. It's called vertical switches, a certain type of switch that, overall, the models tell us that switches themselves aren't large drivers of ignitions when compared to other things like conductor. Conductors create many more opportunities for ignitions than our switches do.
However, when the engineers went out to evaluate these switches, they found evidence of incandescent particles on the poles. So even though the model itself didn't suggest that we should prioritize those, we also listened to the engineer judgment and considered the cost versus benefit of replacing the switches and how many were out there, and we put it into the WMP as a program, even though the RSC might be might look small compared to the other things. So this engineering process, root cause process that we go through on every ignition, provides a lot of value. I think to the broader question you're asking if you're a customer, you know, how do you feel confident? I think the main way you feel confident is that when you read our WMP, it's very clear that we're taking a risk informed approach on everything that we're doing.
So if you live in a high fire area, you're on our radar, because we're putting our resources at the locations where we see the most fire risk.
Speaker 14
Thanks very much. I'll hold off on my other questions, but I would just say, just in terms of feedback, take it for what it's worth. If I lost my home because a sea hook failed, I would wanna read in this wildfire mitigation plan, where and how that key hook was addressed and where it is accounted for in the plan and in the assessment methodology.
Speaker 14
that was, if I'm in that area, I would want to know where that has been incorporated very specifically. Map exactly to where that is within the process and where all of those causes of those catastrophic wildfires have been incorporated into what you do. But with that, I'll I'll leave it there. Thank you very much.
Speaker 6
Yeah, well, can I just, I'll just throw one in, because you brought up the sea hooks, and sea hooks are, what we found, we haven't, a sea hasn't actually had a sea hook fail yet, and cause a fire, but we still went out and studied it? We did aerial inspections, ground inspections, and what we found was that this is another situation where the model wouldn't suggest that sea hooks pose because, a large remember, models are designed using historical data, so if things haven't happened in the past, they're invisible to the model. The models can't predict something that hasn't happened in the past. But what we did do is we went out there and we looked at every c hook in our system, and we realized that because of the way c hooks are designed and where they are, you really can't tell. You can't get a good sense for when they're wearing out.
It's really hard to see visually where they are. But recognizing that, you know, there's evidence that sea hooks can cause fires, We decided to put that into the WMP and we're going change out all our sea hooks.
Speaker 5
Joe, I agree with you on that one as well, and just to, I'll take the pause with Will again, he's referring to something that happened in PG and E system. Sea hooks were something that were not on the radar and because of a catastrophic fire. What we've done now is incorporated that specific piece of equipment into our inspection programs and you can read about that, they are mentioned in our welfare mitigation plan under these inspections, that on, lattice steel type structures that utilise that particular piece of equipment is something that we're now looking at and examining for failure before it occurs.
Speaker 4
Great. Thank you for that question, Will, and thank you for the discussion. Next, let's go to Melissa. Do we have some questions from chat that we would like to bring forward?
Speaker 2
We do. I'll just respond briefly to that to say that the WSD will note that as well in terms of how information gets presented to the WMPs going further, so we'll take note of the comments. The first question that I see is for SCE, and it's from Miles Gordon. A framework for understanding financial impacts of PSPS to customers was mentioned. Could you elaborate on that framework?
And I'll add that if that's better held for the PSPS conversation, we can wait until then.
Speaker 7
So this is Robert, that is part of this presentation. As I said, we've built up the MARS like framework for PSPS so that it could be stacked up with wildfire risk, and I'll just start by saying we put out metrics there in the wildfire mitigation plan, so I'm not saying anything that's not already in the wildfire mitigation plan, and that we intend to evolve this significantly this year as we prepare for both the 2022 wildfire mitigation plan and the 2022 ramp filing that SCE has to file. So as I mentioned, we have safety, reliability, and financial. On the financial side, we used a metric of $250 per customer meter per event, and that of course has been married up with reliability, which is a metric of number of minutes of interruption for each customer based on how many customers are out for the particular amount of time that is predicted, and then safety, which is the metric that is developed both by the access functional needs adder and critical infrastructure adder I mentioned, but also a look back at historic outages, most particularly the Northeast blackout and the safety consequences permitted of interruption from that event.
So hope that helps answer the question. All of this is also described in our WMP.
Speaker 4
Perfect, thanks, Melissa. Next question here is going to come from WSD. Let's see here, questions. All right, so this question is going to go for all three utilities here. So how are different risk models used to different types of initiatives?
So grid hardening versus bench management, how is this implemented and how are the different risks aggregated?
Speaker 5
I'm assuming we're following protocol and PG and E goes first on that one, Alan.
Speaker 5
I think that's a great question and I think it deserves a little bit of discussion here. I'll try and be brief so that I'm not monopolising this conversation and allow my peers to give me their point of view as well. As I described in presentation earlier this morning, and I think we've done quite a good job in our WMP of talking about this as well, we use the risk model to examine historical events and predict future events. And from those predictions we are looking at applying up to date knowledge. So when we build our risk models we're looking at history 2015 to twenty eighteen training data, twenty nineteen test data because this was all done in 2020 as 2020 was going on and then what we want to do is start to pull in what happened in 2020 into that decision making.
So for example, and I know my team later on tomorrow will discuss system hardening and vegetation management in more detail, I'll let them handle the specific questions but at a high level what we're doing at PV and E is we're using the risk model to inform where we should be doing what we call enhanced vegetation management or system hardening projects. To each of those we're adding the reality of what happens in 2020, whether that be rebuilds, circuit modifications that were conducted, fire rebuilds for fires that occurred or PSPS de energisation events and we're including them in the decision as to what the best mitigation is going forward. We've also proven a pool of our team called public safety specialists. These tend to be fire marshals, fire captains, people that have been involved in the fire service at a local or a state level and can help us advise based on what they know of that particular area and the consequences associated if a fire were to start in that area. So we're using the model to inform and reinforcing its practice based on subject matter expertise.
On the vegetation side we do something very similar also. We also use the most recent LiDAR data that we've collected in 2020 and use that to identify strike potential trees, which was a definition that MRT asked for earlier on. That's basically a tree that can fall in or limbs detach and blow in to one of our lines, either distribution or transmission. We identify those trees so the order of execution of some of our vegetation management activities may be slightly different from the order in which is coming out of the risk model to take into account the most recent LiDAR data identifying these hazard trees or strike potential trees.
Speaker 6
I guess I'll go next. I'll three give different flavors of how the risk model influences where we do mitigations. So I'll start with covered conductor. Covered conductor, obviously, it's a fairly expensive mitigation that takes a lot of time for planning, takes a long time to construct, but it has a buys down a lot of risk. It's something that would be done by a crew or a set of crews over the course of weeks or months to cover many miles at a time.
For those, we would use basically straight risk reduction out of the model. So we would rank every segment, and for us, a segment is a piece of section is of a circuit that can
Speaker 6
isolated. Typically, it's between two pieces of equipment. The reason you do that versus, say, span per span and just go after the highest risk spans is because that's not really operationally feasible, just the way you construct it when you build covered conductor, you go through many, many spans before you cut the conductor. So in something like that, we'll follow as close as operationally feasible the ranking of risk by segment. Another example, though, would be something like inspection.
Inspection is something that we're going to, in the accelerated inspections, where we're doing it more frequently than is per compliance, we're going to look at it because it's less expensive per inspection, we're going to look at it in terms of not just straight risk, but really making sure that whether a location is high in consequence or high in probability, or both, that we make sure we capture that every year, because it's less expensive, so you actually get a better chance of going to every asset that could either have high consequence or high probability. So we designed so the values from the risk model go into a matrix, and different areas are broken down based on where they sit in that matrix. So that's a different way they use risk model. A third way would be something like removing hazard trees. That gets a little trickier because in something that you do fairly regularly, you always have to make that decision on can you get more done, even if the risk score isn't perfectly ranked, by grouping things in different areas for operational reasons.
So can you actually, over a period of time, buy down more risk by attacking things in locations versus bouncing around the highest risk tree, then the next highest risk tree, and then the next highest risk tree. So in those cases, you would take the output of the risk model, and then you would sort of figure it out really from an operational perspective. What's the best way to bundle this work over the course of the year, to basically get as many of these trees done as possible, even if, say, early in the year, you're doing, you're removing a tree that has a little bit lower risk than something you would have done later in the year. And then there's other programs that are very important in vegetation, where you do it every year anyway. So any program where we're going to cover the whole territory, there'syou really don't pick up any value by trying to fine tune that year's schedule into highest to lowest risk.
It's better to do it in the most efficient way, which is typically, like for trimming, going straight down a circuit and trimming, trimming, trimming, trimming like that, rather than trying to get so specific where you're going to do this piece of it this week and then that piece of it next week. So hopefully that answered the question. So essentially, and that's why I, you know, what I was saying is why we built our model in a very component fashion, because we realize that every type of mitigation problem out there may need to take a different angle at how to use a model.
Speaker 10
I'll take this one for SDG and E. So just bigger picture, we use the same multi attribute value framework to evaluate all our programs. That's at a high level. And then as I mentioned in our presentation, sometimes we'll come up with more refined methodologies to help us prioritize day to day activities and specific projects within programs. The information coming out of those programmes is fed back into the model.
So for example, with vegetation management, we have to look at a little more detail and specifics around tree species analysis to help us identify what the highest risk areas might be, and we go out and do the inspections and mitigations. Findings from vegetation management, such as identification of trees that have a potential to strike our lines, are fed back into our overall model to assess the segment level risk. Another example is the WAR model that we talked about where specific asset risk is evaluated at an asset level, but then it's also aggregated into this overall segment level analysis. The last thing I wanted to just touch on is just emphasize that we're continuing to evolve and looking for better ways to do this. And I'll just reemphasize what I mentioned about wings being a modular model.
So although we might have specific models applied to address specific problems or help us target specific programs, the idea is that we can feed that information back into WINGS and continue to refine it so that it's taking into account all these data points as we go through them.
Speaker 11
Hi, yes, this is Katie Morsody from Tern. And I have a question for all the utilities. I think as the numbers attached to the wildfire mitigation plans demonstrate there's a lot of potential for rate impacts coming out of the wildfire mitigation plan approaches. And did each of the utilities and how, if you did, include constraints like affordability and limiting cost increases in their risk models as they determined a plan. Hi Katie.
I'm just clarifying, I mean affordability from the ratepayer perspective, not the utility perspective.
Speaker 5
Hi Katie, I think I understand your question there. And I think, and please excuse me, I'm relatively new to PG and E, so if I don't get the mission statement correct and I maybe merge it with another employer that I've worked for, I think you'll find that each of the utilities have a mission that incorporates the word safe, reliable and affordable power for its customers and communities. And I believe that's something that we are standing behind and something that we're trying to make sure we do. However, in my current role as the Electric Operations Risk Management Director, my objective is in buying down risk of wildfire and other public safety related issues associated with the public and our electric assets. So that doesn't mean I'm not focusing on the dollars, but the focus for what we're trying to do here is to reduce consequential ignitions from our system.
Unfortunately it does come with a price. The system hardening activities, whether it be covered conductor, whether it be moving a circuit to a different path to feed a set of communities, comes with a particular cost. And those costs are spread across our entire service territory. So the people in those communities are not being hit with a premium for it. Our objective is public safety and buying down that risk and we are aware that it comes with a price.
We are doing our fiduciary duty to our constituents and our ratepayers to minimise that, but also to focus on making our system safe and reliable for the future.
Speaker 7
This is Robert from SCE. Very similar to Paul's response, I'll just add that obviously affordability is a very important consideration for SCE generally, and there's actually an en banc, I think, coming up in the next couple of days on this exact topic. For us, we do ensure that what we're proposing is in line with our already previously forecasted capital plans and spending plans, and of course everything that we put in our welfare mitigation plan and so it will be reviewed in a subsequent general rate case. All those are, I would say, are constraints that we're focused on, but like Paul, we're primarily focused on coming up with plans that will mitigate risk efficiently and quickly.
Speaker 10
Echoing what Robert mentioned, I think the application of risk spend efficiencies to really inform our future work inherently takes into account the issue of affordability. Because if we didn't have that capability to really target our projects, we might end up casting a broad net of where we want to mitigate things based on the uncertainty. But I think just being able to refine our methodologies really takes that into consideration. We're looking at good projects to implement. And as I mentioned in my presentation, we even go back to see how the new model validates things we've already done.
And we do the efficacy studies to make sure that that's driving us in the right direction. And just like the other utilities mentioned, just echoing the notion of public safety being a priority for us. But again, RFCs are really inherently helping us make sure that the projects we do maintain that affordability.
Speaker 11
I'll hold on my next question until in the chance that I have another chance to answer ask questions.
Speaker 9
Okay, Joseph Mitchell, Mosse Gray Road Alliance. Everybody is training on and analyzing risk events. The question is for data after 2019 and after. PSPS became widespread for all utilities. So the data for ignitions and to some extent outages as well is highly biased and of a different character for 2019 and 2020 versus what it was for 2015 through 2018.
I noticed that PG and E, for instance, is using 2019 as its test data set, whereas it used 2015 through 2018 as its training data set. So this is going to introduce some systematics. How are all utilities adjusting for the bias that's being put into their risk event sample by PSPS?
Speaker 5
Thank you, Joseph, for that question. I think that's a great question in that we have, in our actions, we have changed the landscape looking forward. And as each of the utilities goes through system hardening type activities we're changing the existing landscape. We now have a hardened system. It's something that PG and E is incorporating into the human component but we also have a plan to update our models each year with more recent ignition data as it becomes available.
To talk about your specific point about PSBS, one of the things that PGD has been doing is collecting near net or near hit data with regard to PSPS events. So we had the most recently and this is something I worked on is we had the PSPS outage of January 2021 and we had a number of events on a number of circuits that were de energised for public safety, that then had strikes from vegetation fly ins or equipment damages or what have you, and we collected those and treated those as near hit, near miss type data for future inclusion into our risk models. Because system hardening takes time to be deployed, we are building our data set on that type of failure or those type of failures or near misses, let's say. For the PSPS, we've got a good data set from 2019 and for 2020. I would I would not like to go on record and say that we'd like to get a better data set for 2021.
Our objective is to reduce PSPS events and reduce PSPS duration as well, but definitely by learning from those near hit events and bringing them into our models as we go forward with updates.
Speaker 6
Sorry about that. I wanted a clarification on the question. When you said bias in the model due to PSPS, is that because you're saying that ignitions were avoided during PSPS events, so they wouldn't be captured in an ignition model?
Speaker 9
It's more biased data, not biased model. The data is biased Yeah. Throwing out your outages and ignitions.
Speaker 6
Yeah, think, okay, that's what I was thinking. I thought you meant by it. So for our model, remember, our model is based off of outages that could create a spark. So the number of outages that would take place during a PSPS event, a short period of time on a limited number of circuits, is really dwarfed by the outages that we see across our whole system. So we don't think that it's going to haveit would havewhat we lose in those short time periods during PSPS events, we don't think it would have very much impact on the output of the model in general.
Does that make sense to you?
Speaker 3
Hey, Joe, this is Mason with SDG and E. I obviously appreciate the question. That's something that's been on our minds. For the general audience. Maybe one other way of saying it is, if you're always turning the power off when the wind is above 60 miles an hour, how do you know what's gonna happen to your system if the wind is above 60 miles an hour?
So we're just trying to be smart about how to incorporate that. We obviously collect after action reports after PSPS events to see if there was known or suspected incidents that would have led to an outage or ignition. And we try to take that into account. I will say we're trying to be smart about this. And I don't think we have the exact answer of how to update the models, given that we're working on it as we speak to try to make sure that the bias doesn't lead us down the wrong path, which so often with statistics biases do that.
So, we're just trying to be smart. And in fact, if anybody's got any great ideas, I'd love to hear them and you can share them with the utilities and figure out what's the best way forward.
Speaker 4
Thank you, Joseph. And we just have a few more minutes left here. Henry, I think you'll be asking the last question during this question and answer section.
Speaker 12
Thank you. This is Henry Burton with Public Advocates Office. We probably don't have time for a long discussion, but hopefully we can discuss this briefly. You've all talked about data you use and the methods you use to model wildfire probability and consequences. Can you talk briefly about how you are validating the model outputs to ensure that the models are producing information that's really a reliable guide to real world events.
I know PG and has talked about testing with 2019 data, and one specific question I have there is why twenty nineteen and not also twenty twenty? But more generally, I would like each of the utilities to tell me just why we should have confidence that the models are a reliable guide to real world events?
Speaker 5
Thank you, Henry. I think that's a great question. Let me first of all handle the data question that you asked about, and then I'll go back to the validation. The data that we use to train our models was 2015 to 2018, so that was observed data. And then we run the simulations and after the model was trained and test it on the '29 model year where we had a full year of observed outcomes and we look comparison at of prediction versus actual and develop our area under the curve to test the effectiveness of our modelling.
This was done in year 2020 while 2020 was occurring so we didn't have the benefit of the 2020 data set. What we can do as we go back as we start to give the model more information and we would then test the model say on '15, '16, '17, 18 and '19 against year 2020. 2020 was an exceptional year on its own so we might see some deviations there but again we do have that capability to do so. With regards to verification of the model, because this was a new model, it was something that PG and E took very seriously. We worked with a number of very smart people, PhDs from very good universities, etc, to independently build components of the model and then to test and verify parts of that model.
And also I believe it was a condition of our probation. In RWMP we talk about independent third party validation of a particular modelling approach. The company is called E3. I think it's environmental engineering and I believe it's something that somebody that's familiar with the CPUC. I apologise for not having the full name of their title, but it's something that we committed to in our WMP for third party independent verification of how our model is working.
Speaker 6
And then for Edison, very similar to the way Paul described it, but when we build our models, we use five years of outage data, and the data is broken into test sets and training sets, 80% for building the model and then 20% for testing it, and these data points are selected randomly, so you have, you know, basically a random selection to create each of these buckets. The model is built, and then it's tested against that test set, and there's measurements that machine learning models use to determine how accurate they are, and they give you values like you're familiar with R squared values and regressions and things like that. They have those values as well, which give you a sense to start. Then what we do is we take that model that we built and trained against the test set, and we test it against a full year of data as well that was completely left out of building the model to see how well it's done. Then on top of that, what we do is we now that we've been using the model for a couple years, we can now look at actual events, so actual ignition events.
So remember I was explaining that fire incident process where every time there's ignition, we record in the database, we gather information, engineer goes out there. So what we do is now we have a database full of ignitions, recordables and non recordables, and we've been able now to test what was the probability of an ignition at each of those locations to see where to check that the model is basically directionally correct and identifying the areas that it has the most likelihood.
Speaker 10
So for STG and E, there's a couple of things that we do to validate the models. The first thing that we do, there's kind of a forward looking validation versus a backward looking, and I'll explain each of those. Forward looking is when we construct the model and we have our results, we go through multiple sessions with our SMEs to kind of look through the rankings, look through the areas of concern that the model is pointing us to. And based on their knowledge of the outages, things that are happening on the ground, their real knowledge of where the weather patterns are more severe, they help us validate and calibrate the model. And often we found that we were missing some data points that we then went back and added and cleaned up the model.
Other aspect of backward looking, this really comes into play when we're evaluating the effectiveness of mitigations we haven't applied in the past and may not have a lot of data around whether or not they're effective or to what extent they're effective. So prior to conducting our efficacy studies in 2020, we had estimated effectiveness percentages for traditional hardening, for example. And now that we've had many years under our belt conducting the program, we're able to look back at the ignition rates before and the ignition rates after. And we've actually adjusted our estimates and calibrated the model to take into account those actuals and findings from those studies. Tyson will go over that study in a little bit more detail in the grid hardening panel.
Maisne, anything else to add?
Speaker 4
Great. Thank you so much for all the questions and the discussions that we just had. So now we will be taking lunch. We are going to resume at 01:30PM and next session is going to be vegetation management and Colin Lin will be the moderator for that session. So thank you everyone.
And then the panelists, please let's lower our hands to prepare for the next session. So we will resume at 01:30PM. Thank you everyone.
Speaker 8
Okay, hello everyone. I am reading 01:30, so it is time to resume. My name is Colin Lang. I'm the environmental scientist at the WSD, and today I have the pleasure of moderating 21 WMP updates. Vegetation management can reduce both wildfire ignition and consequence risk.
It involves planning for both the canopy surrounding overhead lines and the understory in and around the right of way. Landscape structures modified by colonization and our changing climate have made vegetation management efforts challenging and require land managers and utilities to constantly adjust to the present, sometimes unprecedented conditions. Over the past few years, all of the utilities present here today have increased the scale and scope of their vegetation management programs to address these issues and many more. In the next hour, we'll hear from PG and E and then SCE and then SDG and E, just the order seems to be the order of the day, about their vegetation management plans. This will be followed by a short break and then a question and answer session panel this afternoon with our presenters as well as key stakeholders and the WSD.
First up, we have Steven Fisher from PG and E. If you are ready, we can take it away.
Speaker 15
Steven Fisher. I'm a director for PG and E responsible for the execution of our veg management program. I actually participated last year as the panelist representing PG and E for this panel. Sorry, so let me get used to this, how this looks.
Speaker 15
There we go? All right, can we go to the first slide of the presentation then, please?
Speaker 15
let me walk through this presentation. So the first part of the presentation is about PG and E's 2021 wildfire distribution risk model, and how we applied that to our enhanced vegetation management program. In 2021, PG and E will be using the 2021 wildfire risk distribution model for its EVM program. We then enhanced this model by overlaying the existing LiDAR data that we have, as well as the completed work from the past two years of our completed EVM program to target the work in the highest risk circuits. This risk model is based on a 100 meter by 100 meter risk pixels that were generated, and each one has a risk score associated to them Due to the way the EVM tool that we use internally is run, those pixels were re aggregated into approximately one kilometer by one kilometer grid.
That's a standard grid used within PG and E for our tracking of our completed EVM work. Any changes to the work plan this year will need to go through a newly stood up wildfire risk governance steering committee. Throughout the year, there might be some needs to change the risk, the targeted circuits that we're working for a variety of operational reasons. Prior to doing so, any changes will need to be approved by this new risk governance steering committee that was stood up at the 2020, and will govern all our work through 2021. The EVM team within PG and E is targeting to complete 80% of the miles from the top 20 of the high risk CPZs, which stands for our circuit protection zones.
Basically the areas where a circuit can be switched on and off. That's how we plan on using the model in our EVM programs here. We'd go to the next slide. Another pretty significant change that we made, that PG and E made this past year, and we'll be seeing the impacts of, is our change to our tree contractor resources and the contract model we use with them. In the past few years, resources have been a major challenge.
There's just not enough qualified tree crews out there to perform the sheer amount of work being performed in the state, not just at PG and E, but I think also at some the other utilities as well. So we've been really focusing on how to improve that workflow or that flow of tree crew resources. So in 2021, we started this in October 2020. We shifted to what we call a defined scope contract model where instead in the past PG and E would identify individual trees and hand it over to the contractors. What that led to were peaks and valleys in their work, which made it difficult for them to maintain a consistent level of staffing.
In this new model, the contractors control the patrol cycle so that they can help to levelize their own workforce throughout the year. So it helps to avoid a lot of those peaks and valleys. It also helps to give the contractor, keep the contractor aligned with what PG and E's goals are, by having them focus on the compliance of the circuit, not just completing so many units. One of the things we saw was as possible in the past that they completed their units, but the work wasn't done successfully and didn't align well. So under this new model, it's more about ensuring that the circuits are compliant with all the applicable rules and regulations, so that if the contractor is truly successful, it also translates to PG and E being successful, and also improve safety overall for the people of California.
So this is going into effect on our routine program primarily. Doesn't have any impact on our EDM scope of work.
Speaker 15
Another area of improvement that PG and E has made in the last year is the rollout of our tree assessment tool. The past, we used a different tool called the hazard tree assessment tool. We've updated that tool. We've made it simpler, easier to understand, And also it's now digital, so that we collect the data as the program is going. So as part of our EDM program, we perform an inventory of all the trees that have strike potential to a PG and E facility, and each tree will get a tree assessment tool, so assessment of it.
So what happens is our pre inspectors are going out, they look at a tree, they're asked a series of questions. If that question results in a severe enough defect that it removal is
Speaker 15
the tool instruct them, excuse me, to remove the tree or to mark it for removal. This tool takes into consideration tree species data based on previous outages. So we are constantly refreshing it with outage data. So as we learn more from what causes and drives outages, the tool is updated as well so that those impacts are seen. So trees that are of a higher risk species get a higher rating for removal than a tree that's less likely to be a cause of an ignition or an outage.
Additionally, what this tool does is take into consideration weather patterns for the area that the work is being done in, specifically wind speed. It looks to what environmental data we have on that area that the person is working in and pulls in wind speed data for that to weigh into the consideration for potential removals of certain types of species. It also takes into consideration terrain and slope, overall tree health, lean species, those kinds of contributing factors all drive this. And so this went into effect in March 2020, and will continue to be used in our enhanced vegetation management program throughout 2021. Next slide, please.
So another area that we've significantly expanded going into 2021 is our enhanced work verification program. So previously, we sent out work verifiers, those are kind of like a quality assurance team, if you will, or a quality control team, to check on the work being performed by our contractors out in the field. We've increased the scope into 2021 for our enhanced vegetation program. Now our inspectors will also fill out their own separate tree assessment tool on all the trees along our corridors that have strike potential. The idea behind this is to make sure that we're getting consistent data on the removal of trees.
It's an area we've gotten a lot of feedback, a lot of questions about Some people are concerned we're removing too many trees, others are not, are thinking we're not removing enough. So by having a second set of eyes or second individual perform the assessment, it'll drive to an increased level of consistency between those programs. Additionally, we're expanding our work verification to our routine program as well. In the past, our routine program had only a statistical sampling done. Now for all the high fire threat districts, we'll be doing 100% work verification of all completed work and all conductors in that area.
Again, with the ultimate goal of improving the overall quality and consistency of the work as we see it across the service territory. Next slide. Okay, so another thing that we realized, and we've had this in place since 2019, but we made some streamlining effects to this, is what we call our priority tag process. Formally, was known as hazard notification procedure. Now we call these priority tags.
And basically there's a priority one and priority two type tags. Priority one tags are reserved for tags that need to be mitigated within twenty four hours. So an example of this would be a tree that showed previous contact with a conductor, or a tree that in the opinion of the individual who's out there performing the work, is in a state of actively failing and could strike our facilities. These tags are done under accelerated timeline. The tags are done within twenty four hours of being identified, and we manage those to completion on a daily basis.
The idea here is to put the highest risk trees or the most dangerous trees up to the front of the queue constantly so that they get worked before they become a danger to the people of California. So it's an accelerated timeline, twenty four hours, and they're identified by the veg management personnel who are out working in the field. We also have a priority two tags. These are tags that need to be mitigated within thirty days. Most common occurrence of these are instances where the vegetation is within that four foot requirement for high fire threats, or 18 inches for other areas.
So if we're inside that compliance zone, we move these to the front of the line as well, so that they get done in accelerated pace. It's also possible for this type of tag to include trees that the inspector has a concern. They may not last until a normal cycle for when our tree crews would get there. And so these are to be done in an accelerated basis, with a similar concept of priority ones with the emphasis on bringing the most dangerous trees to the front of the line so that they're getting worked actively. In the past, we had different timelines for in and out of fire season.
We felt that that created additional confusion, so we just aligned everything to that same timeline, The goal being to make consistent, and the reality is, from our perspective, if we think the tree is in risk of failure, we want to work it sooner rather than later, regardless of whether it's in high fire threat or not. So this is another area of improvement to accelerate the work we do on certain high risk trees. Next slide, please. So another program that we've been working on a little over a year now, and we're going to probably do the most work here in 2021, is around our utility defensible space program. This is very similar to the kind of defensible space that CAL FIRE asked homeowners to do around their houses.
We're looking to do something similar around our power lines, with the idea being that in the event of a failure, the line is more likely to fail into a safe situation. So instead of falling on underbrush, it'll fall onto a more maintained right away with the likelihood of either preventing an ignition at all or limiting the spread or the rate of spread, allowing time for emergency personnel to respond and be able to put out the fire before it turns into a wildfire. As I said, it's mirrored after the same kind of defensible space requirements that you'll see around homes. So we're looking to remove brush fuel underneath the lines, remove ladder fuels adjacent to the lines as well, and also remove some of those higher risk vegetation that we know increase the spread of fire are very likely to ignite in certain situations. I want to note that this is not a complete bare soil removal process.
That's something PG and had tried in the past and met with quite a bit of resistance. It also has some negative impacts from an environmental standpoint. So we're looking at this from what is a sustainable balance, if you will, between going to the extreme of, say, bare mineral soil to a more balanced approach of more like a manicured park style that would allow for still good environmental conditions. One thing we're still looking into, and we are working with our internal environmental team, others outside, would be the application of a flame retardant and or herbicides. The idea with the flame retardant is kind of a proactive approach, prevent the fire from igniting on the vegetation.
We've seen this used quite successfully in various fire situations where flame retardant is applied along poles that are in the potential path. And we've seen that that works quite well. It's very effective for protecting the infrastructure. So we'd like to see the same concept applied here, where the flame retardant is applied on the vegetation under or around the lines, preventing them from igniting in an event where wires are sparking to help prevent those ignitions. We're looking to do some miles here in 2021.
We're gonna model, use the same wildfire distribution risk model that we are using for our enhanced vegetation program to target mileage. The amount of mileage done is still going be relatively small compared to some of our other programs as we continue to test out the feasibility of this program. We're also using this program to work in combination with some local agencies, including CAL FIRE, US Forest Service, and other municipalities for targeted fuel reduction programs in various communities around the state. So that's it for the utility dimensional space. If we could go to the next slide.
Actually, I think that is the last slide. So with that, I think that concludes my presentation for today, and I'll be available for the Q and A portion.
Speaker 8
Thank you, Steven. Next up we have Melanie Jocelyn of Southern California. Melanie, if you are speaking, you are on mute. There we go.
Speaker 17
Camera button work as Technology is not my friend today. But assuming you can hear me now, we can go ahead and, start off and move to the the first content slide, which is really a orientation and review, here in terms of the topics we're gonna review. High level high level review of kind of the major categories of SCE's approach and goals, some key changes, and then we'll go through a few more specific topics in detail. So with that, we can move to the category overview. Okay.
You know, fundamentally, I think we all understand what we're trying to do here in terms of supporting public safety and reliability by managing the vegetation. We're going to really continue on a similar path, expanding some programs and initiatives. And then the specific activities that we do in the field include our hazard tree program, where we're planning to assess between 150,200 trees for 2021 and thereafter. And then we are performing the timely mitigations associated with those, which are typically within one hundred and eighty days once we have the access and authority to remove or perform the mitigation. For our dead and dying tree removal program, we'll continue to do patrols in that space.
And similarly, we have a similar structure for timely mitigations in that space. And then we're planning to do our pole clearing or brushing around the base of between 200,300 distribution poles, working our way up to the entirety of the distribution poll population. We can move to the next slide. Okay. Apologies.
The camera just really will not let me hit that button. So, some key program changes here include our our something that was probably discussed this morning, regarding risk overall, which is the company's move from Reacts to Technosilva. So the way that that takes shape is by leveraging our wildfire risk reduction model where we are applying risk informed decision making. So I am aware, think there was an example earlier regarding, you know, for work that we're doing within the course of a single calendar year, we still consider risk, but it's less necessary to, you know, because you're not excluding anything from the population, the risk factors really come into timing within the course of the year. So where we are not actually looking or working everything within a single calendar year, that would be hazard tree assessments.
Our QC sampling methodology is also dependent on risk modeling as well as supplemental patrols, which are, you know, sometimes called our our summer patrols where we're going, getting one more look at an area before the the time of the year when we have the most heightened risk. Some other things that we're doing this year that we're not entertained at the 2020 timeframe, we're developing an initial tree risk index model. So that will be something that's tied to the WRRM, but we're looking for an opportunity here to apply tree specific information and basically marry up the information in our tree database with the WRRM in a way that will help us better inform if there are opportunities to do inspection, at different cycles or to adjust our trimming distances accordingly. We're doing some work on at risk species. Specifically for 2021 we're focused on palms.
Of the the top species that we have issues with, palms are way up there and so because they often require maintenance three or even four times a year in order to maintain public safety, we're looking to increase the number of removals that we get for those particular species. Something we started in 2020, so it wasn't in the WMP last year, is an focus on quality and engaging with our contractors accordingly. So we're feeding them detailed information on their overall performance. That applies to both inspectors as well as the folks performing the mitigation. And then we're giving you know, having deep dives with them regarding the state of that quality and working with them on corrective actions on how they can do a better job in in either side of the equation.
We're looking to explore continue exploring distribution lidar in a way that would align with trim cycles. So we had an opportunity in 2020, to do some experimentation with distribution lidar and leverage some data that had been collected, but for a variety of reasons, it it was challenging to do the work in accordance with our normal planned work. So we're trying to see if we can come up with a way to advance that area and see if we can do a cost benefit if it makes sense to keep pushing on the distribution lidar side. The final two items are an increase of public and agency engagement regarding notification of trimming practices. That is really to speak to a lot of, customer angst about the work that we do.
And COVID certainly, COVID-nineteen has certainly heightened that experience for us in 2020, and we want to be working more closely with the agencies so that when they hear something from their customers, they have a solid foundation in what we're doing, why we're doing it, how we're doing it, and so that they can help back us up in convincing those customers to do what's best for public safety. The final thing that we're going to be doing is the implementation of our work management system, which we should get through significantly, in 2021. And hopefully, if all goes well, get that implemented for all vegetation work in early twenty twenty two. We move on then. Okay so here's the overall strategy to put everything into context.
Know fundamentally we're trying to reduce or eliminate the risk of vegetation to conduct or contact And so the way that that breaks down is that we want to keep working on achieving enhanced clearance. Based on contractor data, we're getting that about 80% of the time right now. And so we want to keep pushing the envelope on that and doing what we can to work with all of the various constraints that may apply to why we don't have that to date. The main reason for that has to do with maintaining the distance for a full annual cycle. And this feeds into the next one.
Where we can't maintain that clearance, we really want to focus on removals. So these are not changes to the program per se. These are things that we're continuing to push through and try to raise the bar so that we actually are achieving this for more and more trees every year. If we remove a tree that can't maintain cycle for a full year, like a palm for example, that reduces the number of visits we have to make to that customer's property And, you know, hopefully also, even if they're not happy about not having the tree, at least, you know, provides them with some increased customer service just by not having to engage in that process with us multiple times. Finally, we want to remove trees that are fall in and blow in risks, and that is typically characterized by our hazard tree and dead and dying tree programs.
It's not to say that we don't ever remove a tree that would be a falling risk or an at risk species closer to the right of way under our routine program, but these two programs are specifically inspecting for these characteristics where the tree is farther outside the right of way and may not be caught in our routine inspection. Okay next in terms of strategy goes back to that use of the WRRM. And as I already alluded to in the previous slide, what we're trying to do is marry up the tree characteristics that we already have knowledge of with the equipment and risk consequence modeling so that we can put all those pieces together and really drive to where we have an opportunity to adjust our approach in one area versus another. And then on the customer side, we are really looking to, again, speak to that agency engagement, but also some direct customer outreach on evaluating, hey. Is our process good enough?
We know it's not always reaching folks. Is that just because they they really don't believe in in the risk, and is there anything else that we can do to convince them, or are we failing in some capacity to make a compelling argument, and do we need to adjust that? Likewise, we're also looking at expanding SCE's voice of the customer survey process to include vegetation so that we can create a baseline of understanding of how most of our customers view our work. And we can move to the next slide then. Okay.
In terms of inspections, we're doing the following activities. So these are, you know, very, very again, very much kind of factual. We have our routine line clearing. We're looking for encroachments in this case, focusing on distribution in in our high fire areas. That would be anything that encroaches on that four foot minimum clearance.
Then we have pole brushing where we're inspecting the base of the poles. Right? So in line clearing, those trees can be anywhere along the span. It's not tied to a specific pole. So pole brushing, we're actually looking at the equipment itself, and we're clearing the vegetation, really at the base in order to reduce flammable objects underneath in the event of an equipment failure.
Hazard tree really looks, as I already mentioned, further away from the right of way, and it is using a detailed process to score risk from certified arborists in their professional judgment and look to what can we do to mitigate that typically that fall in or blow in risk from large tree limbs or palm prongs. The dead and dying tree program inspects a very specific subset of hazard tree, and this is the long standing one where the tree represents a hazard because it has been, it's so compromised in its health, it's going to die, and that increases the risk of that fall in or blow in. So LIDAR is really our preference for transmission. That's been well encapsulated in our program in terms of using that as a primary reference point for determining what represents an encroachment. So we use the LiDAR to model out maximum sag and sway and clear accordingly on transmission lines.
And then, I already addressed, lidar for distribution, and and how we're trying to evaluate that. Okay, we move to the next slide, please. Finally, on our work management system. So this is a really key effort for us and I think foundational to the improvements that we'll make in ensuing years. We're trying to take numerous digital tools and put them into a single platform.
It's called our Bora. And really there's a lot of benefits to it in terms of the efficiency, ability to apply risk modeling, communication. And I won't read through all of them. Scheduling is a big one. And I'll just pause here just to highlight a little bit.
Trying to manage the work as we do today in different tools, it allows us to get the work done, no question. What it does not do is allow us to optimize it such that we can see all of the various types of work that may be happening in, if you want to imagine, a given neighbourhood block. So if you have a hazard tree that needs to be removed there, maybe you have a dead and dying tree, maybe you have some routine line work, maybe you have an emergent work to address an encroachment. Each of those things today is, maybe a poll to be brushed, right? Each of those things today is in its own system.
And so they are managed simultaneously and linearly. And so we are anticipating the ability to see all of that work in one system will drive a lot of efficiency as well as better data analytics and understanding of what's happening to those trees so that we can, again, use that information to drive the program forward in its maturity. So this platform that our board is based on is cloud based. It orchestrates the processes, automates things in a mobile tool, and then puts everything into a single repository. And so that's also going to help with that engagement with the customer.
I heard about the efficiency, but we also see, you know, it'll be ultimately a better way to able to communicate to a customer. Here's all the things that's happening here, as well as our agencies on the regulatory side, as well as the CPUC, you know, hey, here's all the work in one location. So we're doing this as a phased approach. The idea is to get significantly through the work of implementation by the end of this year and wrap up in the first half of next year. But we are doing it as this phased approach so that we can prove it out for successive programs and functionality before we push to the next one.
Because it's so critical that each, that the whole thing is successful, right? And we need to make sure that our, what, 1,500 plus users in the field, you know, really can embrace and use this tool appropriately. So it's being started in our dead and dying tree program, and we are going to be, pushing that through and, testing it in our hazard tree program next. And this may well be the last slide. Is there one more?
Okay. Thank you.
Speaker 8
Thank you, Melanie. You came in loud and clear the entire presentation, so appreciate you speaking up. Next, last but not least, we have Michael and Tyson from San Diego Gas And Electric.
Speaker 1
Good afternoon and thank you. My name is Michael Delayo, assistant coercer for SDG and E, and I'll be taking you through SDG E's approach, risk mitigation as it pertains to vegetation management. I'll be joined, in the last slide with, Tyson Swetick, who is our Director of Electric Distribution Operations, and he'll give you more detail about one of the analyses that we did in conjunction with our risk mitigation, looking at our outages compared to our. So with that, we're going to get started. Next slide, please.
Just a contextual slide to put us in perspective. SDG, we track about 457,000 inventory trees within our system. And by tracking, we mean that we actually document and keep records of all of those trees every single time a tree is either inspected or trimmed or audited. Rich, robust data set allows us to look historically what's been done and what we may have to do the future to help mitigate any type of virus. Just geographically, we divide our service territory into what we call vegetation management areas, which there are 133 of those.
And we obviously to maintain compliance year round and to avoid any outages or fires, we have to follow very specific master schedule. So this is an activity schedule. All of our activities are contingent on the one that precedes it. So inspection, auditing, trimming, auditing, and pull rushing. Those all together make up all of our vegetation management.
Overall, about 54% of our overhead total system line miles are within the HFTD. That's about 60% of our geographical territories within the HFTD. So we are obviously pretty challenged to ensure that we are mitigating those hazards. We are being on the front end to assuming we're not going to have any trees that are going be contacting power lines. Like the other two utilities, we have engaged trying to adopt and adapt maximum clearances, most from clearances wherever possible.
So we define as enhanced clearances, anything that's greater than 12 feet up to and including a clearance that would safely mitigate any tree regardless of where it is relative to the problem. And then that last graphic is just to give you an example of how many trees in 2020 did we trim and the HFTD and without the HFTD. Next slide, please. Well, with regard to our inspection activities, do have licensing other utilities. We have a specific contractor who performs this function.
We have one train inspection contractor who does our pre inspection activity. We in 2020 have added on four new internal SDGD inspectors. So this is brand new for the company to actually bring on employees who are performing this function. Like our inspectors who do our HWD inspection, these individuals are also certified arborists. So they will be engaged to do all of our, what we call our off cycle inspections in the HFTV and our special patrols include our palm and our bamboo patrols.
So these are trays and along with centric plant, these are species that we have difficulty managing because of their unpredictable nature, growth pattern, and also they just go so fast. But we do actually off cycle inspections in the HFPD for these special cases. Look at our entire HFTD at least twice a year. That's just to ensure that we've looked at the HFTD and if anything's changed since the routine inspection activity, then we can mitigate that before the high persistent fire season. For us, obviously, season is year round, but the most critical time is in the fall through the wintertime, which we have our organic condition.
We're going to continue these post term clearances. We find that is the most effective and our application of that consists of applying this for what we call targeted species. So we have five targeted species that based on historical context, known propensity for species failure, growth potential, and outage history that these trees are the most difficult to manage. So really trying to focus our enhanced clearances that is up to and including those 20 to 25 feet greater, those trees with a known history of causing damage. We also last year integrated an inspection within our HFTD to look at all secondary facilities.
So that includes not only open wire secondary, but also triplex construction. We do not have to drop a whole HFTD portion. And lastly, last year we began an integration with working with some of the local colleges to develop some curriculum specific to line clearance tree trimming contractors and also our crane specialist contractors, we built more professionalism into these careers where folks can actually go to these colleges and get a certification that again could help them with their ability to work in these fields, but also it makes them well positioned to sort of hit the ground running when they start these activities. Next slide, please. So more specifically, how do we track all of these trees?
Again, mentioned we have about 457,000 trees in our inventory. That number changes every day as trees are added and are removed from the system. So this database that we currently have, our work management system, very robust. As I mentioned, every single tree record is mapped electronically. And that tree record houses all of the activity history for that tree.
We know at any given time, for example, what day a tree was inspected, what the clearance was during that day, and the same for trimming and auditing. You can obviously run copious reports on this so we always know the status of our system, if you will, at any given time. All of the data that's captured daily from our contractors is uploaded electronically to a master server. So all of the contractors are in this system, including our tree trimming contractor. In 2020, we started the development of our new work management system, which will create the next generation of our mobile application for our database.
This system is called EPOC and it's going give us improvements in data gathering, work efficiencies. We'll have again the same robust asset history, better report capabilities, and much better mapping capabilities. This will increase the productivity and the efficiency of our contractors and also our ability to capture data that we can use for some of our analytics. We'll also have the ability with this new system to locate by GPS every single tree in our system. Every single inventory tree that we map will have a specific location.
And just a quick definition of what constitutes an inventory tree. It's any tree that has the capability to impact the power lines either by encroachment by growth or could otherwise fall in strike or shed a branch on the power lines. So we'll have information for every one of those trees within the strike zone that meet that criteria. We'll also have more updated customer information and which will improve our ability to engage with our customers. Right now, we have all of our customer information within those tree records for our contractors to use.
And we also have, for example, specific on notification or access instructions. Next slide. So regarding the technologies in 2020, we continued one of our pilot projects with LiDAR. And this is our deep dive into determining whether and how effectively we could utilize LiDAR specifically to integrate it within our routine pre inspection actions, which we are currently looking at that. We piloted a circuit in one of our highest risk circuits where we have relative high tree density and where we've had outages in the past.
One of the drawbacks, if you will, for LiDAR is the frequency with which we get the data. It's subject to our flights. Well, you have to have the LiDAR flight and then you have to produce the data in a fashion that it's usable, readily usable for the department to use. So right now we're currently using LiDAR for similar I think to SCE and PG and E on most of our transmission lines. We can build in PLS pad information and we can derive, for example, the blowout in the SAC of our transmission lines.
We're looking to apply that to some of our distribution circuits as well too, so we can further capitalize on the use of LiDAR. And the other proof of concept issues that we looked at last year in 2020 was satellite imagery. So we're currently doing a use case to look at this technology to see its value. And again, whether it has an incremental value to the vegetation management program. Initially what we find when you compare LiDAR with satellite imagery is that satellite imagery, you can get much more frequent data, right?
You can get this as often as the satellite surface the earth, we could get that data. So it can help us formulate and verify, for example, our tree height data. It doesn't give us the specific spatial accuracy that LiDAR can give us, But we think we will be able to use it in conjunction with LiDAR just to help finesse and give us a more finite look at system. Because required to maintain compliance year round, we really have to rely on our field inspectors not only to assess clearances at a high degree of frequency, but also we need their skill set as arborists to determine not just clearances, but also looking at the tree risks, the hazard that those trees. That requires for us in the HFTD, three sixty degree look at all of those trees within the strike zone.
They could potentially hit the power lines if they were to fall from the ground. Then lastly, we continue to build on our vegetation risk index where we're merging our meteorology data with our tree outage data. Every time we have a tree related outage, our foresters will do a field investigation and will capture all the information relative to that outage, which gives us a rich outage history going for us going back to 2000. And we can mine that data and use that in conjunction with our meteorology history to see where we might have the highest risk. And as Sarah mentioned earlier, with our WINGS model, one of the inputs we can use is the vegetation risk index to see where we may have higher risk on a given circuit or on a segment.
And then we are also, as of last year, we have engaged within vegetation management the use of University of California San Diego supercomputing modeling team. They're actually taking all this data that I've just described and trying to help us build risk analysis and predictive modeling tool. And one of the ways that we're going to help further that along is to capture even more refined data at the tree specific asset level. That one include, for example, what is the soil moisture content for the given area the tree is located on? What is the soil type?
What is the slope? And what is the history of meteorology data? So we're looking at sort of trying all of these technologies to give us the most informed information that could be incorporated and could be integrated into improving our field inspections, which again, we're doing at least twice a year within the HFO. Next slide please. Then another thing that we started last year with vegetation management was really to take a closer look at our sustainability initiative.
And SDG and E, like all the other utilities, is committed to environmental stewardship. Our contractors are trained each year on the rules and regulations as they pertain to environmental compliance. We currently, in regards to our tree removal and tree trimming operations, we remove all of the vegetation material off sites with the exception of large material that we leave for customers. We also upon request will leave material with customers. And then also what we've done in 2020 is we added on a second vendor that processes all material that is brought to their facility into 100% recycling material.
So we're looking for an increase in the amount of green waste that we divert away from our local landfills in 2021. And that little graphic right there, you can see how much tonnage we deliver to recyclable facilities in 2019 compared to 2020. Next slide, please. And with that, I'm going to turn it over to Tyson, who's going to talk about our vegetation clearance analysis.
Speaker 18
Thank you, Michael. I'm Tyson Swedek, the Director of Electric Distribution Operations for SDG and E, and I'll be discussing the vegetation clearance studies that we've been conducting. As you know, we have a goal to reduce risk events and ignitions and vegetation contacts continues to represent one of the higher risks for SDG and E. From 2015 to 2019, we've averaged about two ignitions per year in the HFED due to vegetation contacts, which is second from a risk driver standpoint, only two vehicle contacts. In the 2019 WMP, SDG and E moved forward with a solution to increase post trim clearances based on the significant success we have had historically, reducing vegetation contacts from over 400 per year to under 100 per year when we moved from two to four foot clearances to 10 to 12 foot clearances.
So I teamed up with our vegetation management team to analyse the data from our tree database to see if we could gain some insights into the impacts of clearance from trees to electric conductors on vegetation risk events and emissions. So to accomplish this, we looked at tree trimming data and vegetation contact data from 2002 through 2019. 2002 was selected because this is the year that we consistently started reporting post trend clearance values for inventory trees. We looked at both the vegetation contacts that occurred at a certain post trim clearance distance and the number of trees trimmed in a year at those distances. So this data combined allowed us to determine a contact rate or a number of vegetation risk events over the number of opportunities for a risk event to occur.
We then normalized the data per 1,000 trees trend. We did begin with a system wide look, looking at every vegetation contact on the system over all trees trend. And then in this latest study that we have displayed here, we zoomed in on the specifics of the Enhanced Vegetation Management Program, which is to filter the original data set to just the five targeted species and only trees located within the HFPD. Again, the shape of the curve remains the same and does indicate a relationship that as clearances are increased from trees to lines, we do have less risk events that occur. And I think that's the end of our presentation.
Speaker 8
Thank you, Michael. Thank you, Tyson. Now we have an opportunity for a break before our q and a session. We will resume at let me just double check. I believe it's 02:45.
Yes. 02:45. So we'll see you all back at 02:45.
Speaker 8
I hope that you all got a chance to step outside and maybe look at some trees in preparation for this Q and A. During this Q and A, we will have the speakers who gave their presentations just a few moments ago. We'll be able to hear from stakeholders such as the Utility Reform Network, Massey Grade Road Alliance, the Public Advocates Office, and Green Power Institute. We'll do the same thing that we did this morning. We'll do a round robin.
I will start with one question from WSD, and then we will go around to the various stakeholders and include people who are listening and putting their questions in the chat and the Q and A function in Webex. Ryan Arba will be monitoring the chat. Please make sure when you are submitting a question via chat that you use that drop down menu, and make sure that you select all panelists so all of us can see the chat. The host is in an IT role. Please select all panelists.
Okay. To start out, first of all, thank you for answering the first couple of my questions on my list already. So we'll start in with one a little further down the list. You all all of the utilities have increased their clearances over the past couple of years, in response to increased vegetation management and wildfire mitigation. And a few of you touched upon, fuels management, but I would like to hear a little bit more about how the increased scale of vegetation management has affected, biomass management and fuels management in the field.
Speaker 15
Colin, I'm not sure if you're going to call on us or if it's just going to start with PGE's time, but I'll go ahead and talk about how it's impacted us. So this number fluctuates a fair amount, but I would say somewhere between 7099% of the biowaste generated from our activities goes to regen plants. It depends on a variety of factors, location, also if the plants are even taking anything. Unfortunately, what we saw in 2020 with the large number of fires that occurred, many of the plants were overwhelmed because there was just so much available material for them to be taking, and so they got to a place. We do tend to work though, even if it doesn't go to one of those plants, the landfills we go to, we make sure that they are, that they do treat the wood environmentally friendly means, that they don't just let it build up in places.
So we try and work with folks to make sure that it gets to a reasonable place. But we have seen that it has overwhelmed at times some of the systems that are available to us, specifically with the wood that we haul off. We tend to only haul off wood that's less than four inches in diameter, and the larger diameter wood we tend to leave in location. That can vary depending on the situation pretty drastically, but that's our general policy. We remove that smaller diameter material from most locations that we do work on.
From a fuel reduction standpoint, a targeted fuel reduction, as I briefly mentioned in my presentation, over the last few years we've done some targeted fuel reduction projects. Usually those are in partnership with a local agency, typically to establish fuel breaks around cities or other critical locations. In 2021, we're going to be targeting a more a larger fuel defensible space program that I mentioned, and that'll be along our right of ways.
Speaker 8
Quick follow-up question for you, Steven. Does PG and E have any programs to help increase the number of pathways for biomass waste when the system is overwhelmed? Or do you see that as more the overwhelming volume of biomass as more of a temporary problem?
Speaker 15
So the first time we really saw that issue was in 2020, and that was really over 4,000,000 acres burnt in the state. Prior to that, we haven't really seen that at all. I don't know of any plans to increase that, that PG and is involved with right now, but we do work and support those different agencies often. I think the majority of our waste we give to them is at no cost. We don't usually ask for anything in return for it in most instances.
So we try and make it as beneficial to them as we can in our practice of giving it to them.
Speaker 17
Hello. So I can address this one as well. So in some ways, we're similar to what Stephen described. Basically, except we don't distinguish on debris size. The work that we do is generally always hauled away and disposed of and or or recycled by our contractors.
We don't track closely the amount that's recycled. But typically, from a financial perspective, as long as there is a center in the vicinity that is willing to accept the debris that recycle recycling simply because it's affordable that way. And the only tree debris that I'm aware of that cannot be recycled is those from palms because it's not actually a a tree. And so, as I mentioned, the debris of whatever size is removed. There are certain exceptions in that case.
The major exception is remote areas on US Forest Service land where the rules permit it where more than a 100 feet from the roadways will cut and then we scatter it up to a certain height so it doesn't create too much of a fire concern there. We are actually also undertaking a fuel study with the Edison Power Research Institute through the remainder of this year and hoping to get some further insight there on best practices. And then just to anticipate your question in terms of biomass, we are aware that again, terms of use of biomass plants for energy, that is also sometimes an option in the vicinity of where the work is taking place. But part of the challenge there and our ability to support that effort, which is something we've looked at, has to do with the continuity of the work, Meaning if somebody is going to set up a biomass generation plant, they need to be able to assure a consistent fuel supply. And that's not the way that we plan our work, right?
That is not our core purpose. And so it tends to be based on economic factors for the region and something that we can take advantage of rather than something that we can explicitly pursue or drive independently.
Speaker 8
Thank you, Melanie. Can we get a response from Michael or Tyson of SDGME? There you are, Michael.
Speaker 1
Yes. Hi, Michael. Thank you. Thank you for that question too. So I would just reiterate for our routine vegetation management operations, we would shift all the way, almost the exception of the larger degree.
The larger degree would be, for example, if we're taking large diameter trees down, that is customers properties that we do chop it up and leave it into manageable lengths up on the property. When it comes to agency lands like Forest Service or State Park, we follow their specific guidance and requirements for free disposal. So as Melanie mentioned, for example, Forest Service, we would loft and scatter. There are some communities where we have entered an agreement to remove most of the material, even the larger material off-site in conjunction with our routine removal operations. But majority of the debris is chipped and off-site.
As I mentioned, last year, we were able to divert about 4,500 tons of material that would otherwise have gone to landfill into a recycle facility. Prior to our fuels management program, one of the initiatives that we continued last year was what we called our fuels modification treatment programs, where we're reducing fuels around the base of poles to try to prevent ignition if there were to be either an equipment failure or a pole failure. And what we're doing in those instances is we're removing all of the flashy finer fuels out 50 feet from the pulp to avoid an ignition. And those in those instances, all that material that we generated, was actually donated to parties that wanted to use that ground cover or whatnot.
Speaker 8
Thank you, Michael. For our next question, we'll go to the rest of the panelists. Marcel from Tern, you want to unmute yourself and ask your question.
Speaker 19
Okay. Sorry about that. Thank you very much. This is Marcel Haber from Tern. I have two questions really about benchmarking for
Speaker 19
the utilities. So the first is I think all of you described that you have some sort of program to remove what you might be calling hazard trees or target trees or target tree species. And I'm just wondering if you can if you have if the utilities have gotten together to discuss the criteria they use to identify such trees and whether you've done any benchmarking concerning your hazard tree programs. And if so, what's the outcome of that? Let's see, may I ask my second benchmarking question right now?
All right. I'm going to take that Yes, as a yes
Speaker 19
Well, it's a different it's benchmarking, but a different topic in vegetation management, and that is you also all described sort of systems you're implementing or have used for work management. And I know the whole issue of quality assurance, quality control is a very kind of critical issue in the area of vegetation management. And I'm just curious, it sounded like Edison was describing the Orbora mobile tool and SDG and E described its EPIC tool. And I didn't quite understand whether PG and E's statement that is moving to defined scope contracting constitutes something similar in terms of work management. I guess what I'm really trying to ask is PG and E sounded like it's going a different direction from internalizing work management through centralized databases to kind of outsourcing it a little bit more to the contractors.
Let me know if that's totally if I'm misunderstanding. And my question is whether you all have benchmarked any of these work management tools. Do they I'm assuming they include quality assurance, quality control. But please tell me if I'm wrong. Thank you very much.
Speaker 17
I'm happy to start. So I can tell you that in terms of hazard tree discussions, it's been some months since we got together, but all three of the IOUs represented on this call have had benchmarking discussions regarding hazard tree going back to at least early twenty twenty. And my memory is failing as to what about extend to 2019. And I believe that our fundamental approach is the same in terms of evaluating the risk characteristics of each individual tree and then acting accordingly associated with that tree. But rather than, you know, letting it stop there, I'll let my colleagues weigh in on that point.
Speaker 1
This is Michael with SDG and E. I would concur with what Melanie just said. Thank you, Marcel, for that, both those questions. Regarding hazard trees, so just to reiterate that the individuals that do our hazard tree inspection within the HFTD, and again, we do two separate inspections within the HFTD and these folks are qualified to perform these assessments. That is they are certified arborists.
So they're trained and they're educated in, for example, tree growth, tree failure characteristics, which trees have the propensity to fail. But I think, well, there's some commonality with all the utilities. Obviously, you know, we live in different parts of the state. There are some common species and common attributes that are required to assess when you're doing hazard tree violation. We are a little unique in that we do not have nearly size tree density, nor species diversity.
We have, what we call five targeted tree species that again are known to grow fast or historically have caused the highest frequency outages. And it's those trees that really focusing on our enhanced operation function with hazard tree evaluation. By the way, those species are oak, mark, pine, palm, and eucalyptus. Now the eucalyptus, it's non native to California. There's hundreds of species of eucalyptus trees.
We unfortunately have a few that are notorious for failing. They grow fast, but they are also what's called a top wood species. So they can fail during, especially during windy conditions or when the soils are saturated. We were also again looking at the frequency of our outages to help inform us and drive those trees that we're really trying to practically mitigate. It really comes down to the completeness and the accuracy of hazard tree assessments and the individual that is trained to do that, where you're going to see the highest success in identifying hazard rates.
Lidar and other technologies can inform you, but it really requires those people around find their professional knowledge. As it relates to the second question, it's all about work management, go into more detail about our specific system, what we're striving to do is track our trees at the asset level. So we know at any given time, again, when a tree has been inspected and how much appointments it had at that time. So it could better inform us what actions to take. And we do have a specific third party quality audit activity that takes place at the conclusion of all of our vegetation management actions, inspection, and.
Speaker 15
I'm not sure, did we lose Michael? I'm not sure if others can still hear him.
Speaker 15
All right, well, I'll just add on. As you heard, we have met together to talk about different trees, targeted tree species rules, but again, it's not so much the species. Like, for example, both you heard Edison and San Diego mention palm trees. I don't really worry too much about palm trees up here in the northern half of the state, so it's not a major concern. Gray pines are a tree that I have a lot of concern with, and while I wish we were doing some targeted species removals, that's not the approach we've taken.
What we've done is we've developed our tree assessment tool, and what that tool does is takes into consideration outage data, as well as other data, as I mentioned earlier, I won't rehash it here, but basically looking for factors. What are factors that make a tree more likely to fail than the tree adjacent to it would be likely to fail under certain conditions. And so that's what we look at, and we've met with the other utilities and talked about what are some characteristics that we should be looking for that help drive those discussions. And I think as you heard, we also all use our outage data. Where we've had failures in the past have shown give us an indication of where we'll have failures in the future, both in location but also in tree species.
And then as for your second question about the technology we use, so I think like the other utilities, we've been looking at a new tool to oversee our veg management program. PG and E is also undergoing that same process, hoping to have something in place for 2021. To your question specifically about defined scope, we are looking to take a step back and not be in the way, if you will, of our contractors and their ability to be efficient. In our old model, our pre inspectors would give us the work and then we would turn around and give it to the tree contractors. That created a middleman step.
So what we're looking into with our defined scope, which is limited just to our routine program, is that the tree contractors are working directly with the pre inspectors so that they control that flow of work. And we started this in October, and already we've seen improvements in that, especially that communication between the two groups, the tree contractors and the pre inspectors, allowing them to be more efficient. But in terms of the data that we collect, PG and E will still have all the data, will still be the repository for that data, and so we are in the process of developing a work management tool as well as a database, a new one to replace some of our older systems that all those different groups can use to store their data across that we contract with.
Speaker 8
Thank you, Steven. Just a quick clarifying question for you. I think you mentioned that you may not target specific tree species, but correct me if I'm wrong, the tree assessment tool does include species as a risk attribute.
Speaker 15
It is a risk attribute, but it's not an automatic, right? So I mentioned, let's just go with the grapevine that I mentioned. Just because you're a grapevine doesn't mean you're automatically going to be targeted for removal, but the scoring system that works in the backend of that tool will make it more likely that that tree itself will be targeted for removal, but it's not an automatic.
Speaker 17
I just wanted to make that distinction. I think one tree that's probably consistent amongst all three utilities areas is the eucalyptus. So similar to what Steven described, we're not going to unilaterally remove all eucalyptus. But because we know that eucalyptus tend to shed limbs more easily, it's more likely that a eucalyptus would reach a threshold that requires mitigation. It's going to depend on how far it is from the line, a direction of lean, various health factors that are based.
And that's where I started with in terms of doing an individual particular tree. And then and then just because I didn't address Marcel's second question initially, I just wanted to add there that, both the tool that we use today and the Arbora tool that we're working to implement do incorporate QC as part of the functionality. And we have had extensive benchmarking with our colleagues in discussing options before we pursue the path of that tool.
Speaker 8
Thank you for that clarification, Melanie. Next on my list is Will Abrams.
Speaker 14
Thank you. I just had a question on some perceived disconnects with the vegetation management approach. One of these was touched on in the morning part of the of the discussion around the bow tie methodology. So part of what I see sort of missing there on the far right hand side of that bow tie are the mitigation efforts. So specifically around vegetation management, you know, wildfire occurs, and then we have MTTR dictating how quickly we can restore power, right?
And so part of what I don't see in the plans, and maybe you can speak to this, is here's how we're here are the plans to address vegetation management post catastrophic wildfire. Here's where we intend to underground in this high fire threat district. Here's where we intend to do fuel breaks, fire breaks, when that limited window of opportunity arises post fire, where you can do some things that are different, rather than just sort of putting lines right back where they are. Similarly, you know, a disconnect with the other wildfire mitigation plans, or community wildfire protection plans that exist. So what I don't see is sort of a systemic or integrated way, where the wildfire mitigation plan is talking to the community wildfire protection plans and other statewide efforts like Governor Newsom's 30 by 30 effort to restore the wildlands.
And seeing that integration, because obviously, these plans don't exist in a vacuum. And while I've heard some sort of anecdotal, hey, we work with this city when they reach out to us, I wanted to see where, as part of the planning effort, these things are more systematically incorporated. And the last point in terms of a disconnect is really trying to understand what are the external variables and what do you have as sort of a plan B, plan C to respond to that? So those things might be things like COVID, where how are you going to address depending on that? Or if you, you know, your tree trimmers or contractors, you sort of stop their tree trimming and you have to find other available contractors.
Just trying to understand where you have built into those plans, that type of integration, and where you can fill those gaps. Thank you.
Speaker 15
Well, I'll take a stab at trying to answer that. So first off, let me start by, I don't know that I necessarily agree that I see a benefit in a post fight or situation to do things different than what we're currently doing, and the reason I say that is, is it's seldom in a situation after a fire that there's no customers left, right? What my experience has been over the last few years is that while there's a considerable amount of destruction, there's still some surviving homes and people in those communities who want their power restored, and are not always able to wait for, say, an undergrounding project to occur or the lines to be relocated. And so that is something though that we do look into. I know PG and E has worked to identify situations where it makes sense to underground, but undergrounding is a very long process to undergo, and it's when you still have surviving houses and customers who want power and need power to resume their lives, being able to stop and just do undergrounding is not always a viable solution.
We do have some situations where we've done alternate generation means, so I know we have quite a few locations where it's basically, we've got miles of line feeding a single customer, and we use these opportunities both after a wildfire or during our normal course of business to target those individuals and either convert them to solar or convert them to some other means of generation so that they can get off of the normal grid, and then we don't to continue, we don't have to rebuild the line or we don't have to maintain that existing line. And then for your second question, and I wasn't totally clear on this, so you're asking about how we have contingency plans for the unexpected things that can occur, like such as COVID, was that where you were going with that?
Speaker 14
Yeah, so the other two areas are integration with other wildfire mitigation plans, like community wildfire protection plans. So as an example, I was on the steering committee for the Santa Rosa, and the county has another community wildfire protection plan. And what I don't see within those and others is here's where PG and E or other utilities tie into those plans. That's where your vegetation management is within their strategic plan, and their efforts are embedded with your plan. And yes, those external variables are important because part of what I gather from what you said, which is a disconnect for me, is every wildfire professional that I've spoken to, is that there is this unique opportunity when a wildfire hits, to turn a burn scar into a firebreak or a fuel break, or to think about undergrounding.
I think PG and E actually looked at the situation in paradise as an example, and said, look, that's where we need to underground, sort of learn from their mistakes in 2017, with just putting it right back where it was, even though there's a lot of pressure there, to have that strategy in place, so that you're ready to go. So just trying to understand what you've developed since that time to have it more ingrained in your wildfire mitigation plan, if that makes sense.
Speaker 15
Right, well as I said, we do try and target areas for another approach in view, but I would say, one of the other pieces is that the utility's responsibility is to its infrastructure. That's our core responsibility there. And of course, we want to be partners with the communities, and that's why we do work with the different communities and what they're trying to accomplish. And you'll see that there are examples where PG and right of ways, especially our transmission right of ways, are used as natural fuel breaks, or not natural, but newly created fuel breaks to protect certain cities, but doesn't always work out that way, and so the location may not be ideal for a fuel break, or may not be included in the local community's plan for where the established break is, and so while we're certainly willing to work locally, there are additional steps and requirements that kind of make that difficult, but I'd like to think, and if this isn't the case, I certainly will take this back, but I'd like to think that PG and E is always willing to be the table during those discussions to be part of that.
As I mentioned in our defensible space presentation earlier, we are working with those communities when they do reach out, and we're happy to be at the table during those discussions and look for opportunities to collaborate, but my job is to make sure that the overhead lines are clear of vegetation, and when I can find opportunities to overlap with local community, I'm more than happy to do it, but I'm not gonna create a fire break that's outside of my easements. That's just not what my responsibility is as a utility, but I'm happy to partner when they do line up well. I don't know if any of the other folks have anything
Speaker 17
So in terms of speaking on fire restoration specifically, I think Steven raised a couple of good points in terms of the urgency of restoring power in the moment. For SCE, we had a significant fire in our territory last year, which was the Creek Fire and required over a thousand vegetation management personnel, mostly contractor in order to support the restoration activities up there. So, you know, certainly when the moment arrives and we're trying to clear trees that have been damaged from the fire, we're trying to protect the, not just the infrastructure that's going to be built and create enough room to build the infrastructure, But we're trying to actually create the space around the infrastructure that we need to maintain on an ongoing basis. And I make that distinction because I don't want folks on this call to walk away with the idea that it's always really easy to keep that clearance, right? We have a lot of barriers that come up from the people who own the property, who own the trees in many cases.
And so what we can do in a situation of fire restoration is talk to the property owners to the extent that they're there or agency, if it's a federal or state agency, local government land, and talk to them about expanding the opportunity for us to create that full clearance. And in the process, that may well create a firebreak. So I think, you know, it in fact an opportunity that we're seizing on to say, okay, not only do we want to get, you know, just get the facility, you know, get the trucks and the equipment in and out of here, but no, we actually want to set this up in a sustainable manner. And I can tell you from the Creek Fire, one of the other things we're doing is we're using this as an opportunity to do a pilot on some integrated veg management where we're So at a very small it's not highlighted in the WMP, but we're looking for opportunity to see if we can foster the growth of more low growing species, which are more compatible with our right of ways and not something that would require maintenance in future. In terms of integrating with or dealing with contingency plans in terms of local communities and their planning, the major area of intersection for us seems to be with leveraging local fire councils to support our efforts.
And so while it may not be integrated at the plan level, and I think that that's something worth pursuing, because if we can integrate our work into their plans, that would certainly bolster our position in making the argument that we need to trim X tree, remove Y tree. And so I will take that back as something we can consider to formalize that. Right now, what we tend to do today is when we're having conversations with a community, let's say it's a city mayor or board or dealing with individual homeowners associations, we'll try to generate support from that local fire chief or the fire agency and just get them to understand what our plan is and how that does integrate with their desires. And in some cases that is effective at overcoming resistance regarding the work that we're trying to accomplish and maybe not always as easily as we would like. And I think one final point you mentioned was around COVID-nineteen and contingency.
So I will just add there, Will, that the internally, we have a pretty strong matrix approach to see what happens if we're unable to complete all of the work we have planned under those circumstances. And what are the thresholds that would require us to reconsider various actions under those circumstances?
Speaker 18
Yeah, and just to add a few more things to this. I know PG and E and Edison have answered it very well, I would just say too on kind of the context from can you like go in and underground like a whole system on something that burned down afterwards? Our offer mitigation plans are very focused on preventing the catastrophic ignition. Preventing all risk events at ignitions. But I would agree with what's been said.
There's usually still some customers that are that, you know, do have homes that are not coming down and we're trying to restore service as quickly as possible. And to restore service, there's a lot of engineering involved no matter what you're doing. A lot of the reason restoration is fast is because we're going back with the exact same structure, types, locations so that we know in attachments, it was there before so that you're so that you're, you know, the engineering's already been done to know that that's going to work, right? You have the right clearances, you have the right heights, you have the right structure loading. If you were to say, hey, we're just going to go ahead and underground this, well, you haven't done any conflict studies.
You don't know if you're going hit water pipes, gas lines, all the things that are going under the ground. It takes, you know, about a year typically to do a full design. You need to get your cut lengths, your cable pulling tensions, siding your manholes. There's a lot I would say there's a lot that goes into planning the design, engineering, construction of a line to be reliable over the long term. And then when we get into emergency planning, we really are typically just going back with what was there before because we know there's a design that worked reliably before.
So, and again, I think in the plan, the focus really is on prevention. That is the goal, to not start those fires. Thanks.
Speaker 14
Thanks for that answer. Just guess what just something to take back. It really is this unique window of opportunity to do things at a much lower cost. You know, roads are already gonna be rebuilt. People's homes are destroyed, you can underground more easily.
There's opportunities there. So some of the things that you just talked about in terms of doing that strategic work upfront, so that you have a plan in higher fire threat districts, that you can both do the short term restore power and improve the longer term mitigation efforts. Think, seems to me like an area that we should all be focused on, not just the utilities, but all of us coordinating together so that we can use these opportunities to really work together towards common interests. But thank you.
Speaker 8
Thank you Will for your question. Next, Zoe Harold from the Green Power Institute.
Speaker 13
Hello everyone, Zoe Harold from Green Power Institute. So our question is essentially a continuation of the WSD question on fuel load and then VM and EVM residues. And I'll ask it in two parts. There was a mention that Palm has limited use pathways, but we know some biomass facilities, for example, the 50 megawatt Desert View facility in Mecca can actually utilise palm residues. And so our question was, have the utilities with palm residues actually explore these biomass facilities and residue pathways?
And then the second part of that was for PG and E. So we are curious as to whether the residue removal process and incomplete removal of those residues is leading to the need and consideration of using fire retardants. And so if there was more complete removal of those residues, would that reduce the need to use fire retardants? Thanks.
Speaker 15
Okay, well, I'll answer. So PALM's make up such a small percentage of our trims, I don't think we've looked into that at all. Don't even think it's, but I'll follow-up on that, but I don't think there's much in the PAMS. But specifically to the reason behind the fire retardant, again, it goes back to prevention. It doesn't have anything to do with the loading of the fuel.
What we're looking to do is you can have a wired down event for a variety of reasons, you know, a bird strike, equipment failure, a vegetation strike, car pull, the list goes on. And in all cases, what typically causes the fire to spread is that flashy fuel, that brush that's underneath the conductors. I've been out to multiple sites where a tree has fell on the line, and the tree, which is still alive, tends to not burn, but it's all the dead brush on the ground. And so the concept behind applying that flame retardant would be that if the wire does break and has a downed event, it doesn't have anything on the ground that can ignite. And we've tried in the past, PG and E's tried in the past of going to bare mineral soil that met with some pretty stiff resistance from the communities, and also just, has issues with sustainability.
But the flame retardant is something that we're looking into because you can still have some vegetation on the ground with that flame retardant applied to it, and in some of our laboratory tests that we've done, it reduces the likelihood pretty substantially of an ignition breaking out. And so, we're looking to do prevention, that's our primary focus here, and what we've seen just in laboratory tests is that the application of that flame retardant significantly reduces the likelihood of a spark breaking out, turning into a fire, and then it also has the added benefit that it prevents that, or slows that fire from turning into a wildfire, and that spread is so much reduced that local emergency personnel can get out there and put the fire out before it gets out of control. So that's why we're doing it. It doesn't have anything to do with the fuel loading left by our work or by other existing vegetation that's out there.
Speaker 1
And this is Michael with SDG A side, just to concur with PG and A. So similarly, we tried a pilot project, last year with the use of, Fortify, which is a top of the farm product name. And we piloted on, about a two mile linear stretch of our facilities, about 100 mile path. And we were just looking to test and see again, if incrementally what added benefit can we get as, to reduce fire ignition. What we're finding out though is that, it comes at a very high cost, especially when you look at comparison between traditional cold rushing, where you're mechanically removing all of the vegetation.
So we're going to continue, looking at that, the use of that product this year, again, to make a comparative analysis, whether it makes sense to continue this from an efficacy standpoint and, an efficiency standpoint. Where we differ with PG and E obviously is with regard to palms.
Speaker 1
San Diego obviously, the premier signature species, most of the palms that we have are non, endemic, they don't grow here and, or they don't originate here in San Diego. So we are really challenged with, palms as a species. That's one of its, it's one of our top five species. So really proactively trying to remove palms whenever we can. We do that with, very robust engagement with customers, explain to them the reasons why it's beneficial to take the palm out.
And, unlike other species, palm is not even a tree species, it's a member of the grass family, we cannot direct its growth, away from the power line. So once it reaches the height, we have to remove it completely. And then we're left with, the residual. A lot of our, most of our landfills do take palm debris, that requires a higher tipping. Our recycling facility, one of our recycling facilities does take pumps, but it is of a limited use.
So there is not, as much of a higher market, if you will, as a recyclable material. But, we are interested in investigating different options for utilization of palm materials, especially given that we do generate a relatively high number of palm removals every year.
Speaker 17
off with Palms as well, just going on for Michael's response. So I think we've heard in the prior two responses some reference to economics and the fact that we are stewards of our rate payers and need to keep rates affordable for them. So I heard you on the location in Mecca that would be interested in palm waste. So I think to the extent that we have palms in that vicinity of our territory, at least leading out in that direction, because that is outside of, I think, all three of the utilities territories where it becomes economically feasible to transport to that location, that definitely seems feasible. I think what we would find is that in a situation where many of our palms are actually located in coastal areas, so up near Santa Barbara for us, and so if you're contrasting option to take that to a local landfill that would accept the palm versus a much farther interior location that might have a higher and better use, it would come down to is the economic cost or additional rate that that would create for the customers a benefit for which the environmental benefit outweighs that.
And so I think that is something that we take very seriously. We need to look carefully at what the impacts would be for extensive transit of that kind of waste material. And I don't know that there's a lot more to add with regard to the idea of the product Fortify, which Michael specifically identified. But we are also taking a look at that and trying to see if that in any way fits into our future.
Speaker 8
Thank you for that question, Zoe. And I have a quick follow-up question for Michael and Steven. You both talked about piloting fire retardants in your right of ways. And I was just curious about the length of efficacy of that fire retardant and whether, like, when it's applied and how often it needs to be reapplied to be effective.
Speaker 15
So in our initial laboratory tests, we actually haven't done much out in the field aside from doing a fire response, but during our laboratory tests, we were told that basically it's good for a year, so it would have to be reapplied annually, or after, I believe it was two inches of rain, which if we got two inches of rain, we would be extremely happy and wouldn't be worried too much anymore about the fire retardant. So that seemed to, I mean, it'd be nice if it was better than annually, but seems to, it seemed like it would last during fire season is our initial reaction.
Speaker 8
I think we'd all rather it rain than it be effective for multiple fire seasons.
Speaker 15
Sorry, one other thing I was add is we did also see a drop off in efficiency for areas that were prone to high wind. It doesn't go to zero, but it does have an impact. Sustained wind, which again was done in the laboratory, did seem to have an impact. That was the only factor we noticed.
Speaker 1
Oh no, no worries. Thank you, Stephen. Yes, similar for SDG and A, discussed with a pilot project, relatively small, area that we treated. Looks like we can, we can get one, maybe two years out of this, but it all depends on environmental factors like, the amount of precipitation. At this juncture, we don't think it's going to be necessarily a replacement activity.
It would be an augmenting activity in conjunction with our rushing, our enhanced vegetation clearances, and reducing some of those finer fuels at specific holes that we have within our region.
Speaker 8
Great. Thank you, Paul. Next, I wanna move to the chat. Ryan, are there any lingering questions in the chat?
Speaker 16
Hey, Colin. Yep. We've got two questions here in the chat. First one for Steven in PG and E. Is the tree assessment tool a proprietary program, or is it based on the ArcGIS Collector app?
Where does it pull the wind gust data from? Does this have to be entered manually by the inspector?
Speaker 15
So the tool itself is actually, we use Survey123, which is an add on to the ArcGIS Collector app, so the program itself, I guess is, I mean, it's not PGD proprietary, you'd have to have access to ArcGIS to use it, which we bought a license for. It's more the questions, a series of questions, which I try to get an answer on this, and I think the general response from PG and is it's not proprietary, it's just a series of questions, and I believe we probably disclosed it certainly to some of the regulatory agencies, the series of questions. The wind data comes from PGE's meteorology team, and they don't have to enter it, the way the system is set up is that when they create the inventory point, when they're out in the field, it takes a GPS coordinate of where they're at, and pulls the relevant wind data. It's not real time wind, it's average wind speed for that area that they're in, so that that is a factor. So on the simplest of terms, it basically would put in an added factor, says this area gets high wind speeds, which would increase the likelihood of a failure, which would increase the likelihood of the tree needing to be, or being removed if it has additional defects that might make it weakened in a wind type event.
So I think that's all the questions for that one.
Speaker 16
Got it. All right. Thank you. And Collin, there's one more for Edison. If you'd like, I could read that one off now.
Yes, go ahead. Great. For decision 20O5051, utilities are supposed to provide a customer friendly portal that allows the public to view clear and comprehensive information on their specific vegetation management activities that are ongoing, the progress that has been made, and the expected completion dates. SCE's website just provides a general statement that 1,100,000 trees are inspected annually and 750,000 are pruned. From the presentation today, it appears that SCE is able to provide the type of location specific comprehensive vegetation management information that is called for by decision twenty five thousand and fifty one.
So when is that information going to be made available on a customer friendly portal?
Speaker 17
I don't have an immediate information on date of availability. I think we're going to need to see the Arbora tool that we've been discussing implemented in order to translate that data to something that would be a comprehensive view for the customer base. And so we would need to be looking ahead to 2022 for that.
Speaker 8
Thank you, Ryan. Next, we will move to Joseph Mitchell of Boston Great Road Alliance.
Speaker 9
Hi, this is Jess Mitchell. I had a question on the classifications of at risk species. This is for all the utilities. In the past WMPs, at risk species were determined as far as outage data informed at risk species determination, just by the sheer number of outages. And as a result, the top at risk trees tended to be just the top trees in the population, rather than normalizing as the tree frequency in the utility area.
I know that we did some analysis on some of the quarterly updates with SDG and E and saw that some of the weights for some of these at risk species were greatly different from each other. Palm, of course, always being really bad. But certain things not making the top of the list, like cypress and century plants, just because there aren't that many of them out there. The question is, as you're maturing these risk determination models, are you, and you're incorporating outage information, are you normalising that outage information per tree, per outages per thousand trees per year, for instance? Or are you still using just the sheer number of outages from your service areas?
Speaker 1
Well, this is Michael. I'll go ahead and kick this one off. Hi, Joe, thank you for the question. So I'll, I'll reiterate just for context, our, our top five species, sycamore, pine, eucalyptus, prong, and oak. And Joe brings up a valid point about, there's the relative frequency that these trees cause outages.
And then there is the specific tree's species propensity to cause an outage. That can certainly change year to year. I think what's important and critical for us is, every time that we have an outage, we do a very detailed analysis and investigation of not only the species, but what were the mechanics that caused that outage? If it was a branch breakout, was there, for example, included bark? Was there a severe bleed?
Are there other environmental factors that could lead us to thinking could have been a contributing factor to that being an outage. Even though we consider those five species to be our top five species, there are, for example, some species where we're going to nuance it a little bit further, and I'll illustrate that further by bringing up pine species. So, SDG and E does not have a very large diversity or number of different pine species. But there are a couple of pine species that are especially problematic for us. They are our non native, I'm sorry, pine species.
The Aleppo pine and the Canary Island pine. Most of our outages that are caused by pines are one of those two species. Found that they have a propensity for failure. We ironically do not find a lot of outages, for example, in our higher elevations where pine species are usually found. So one way that we are getting into a little more detail in the application of our targeted species and where we in turn get our greatest clearances is to focus those greater clearances on those specific three species.
That is the Aleppo pine and the Canary Island pine. Similarly for oak, we like most of California have a lot of oak species nowhere near what our counterparts have up north. There are a couple species that are, for example, are, subject to, pest invasion. We have a pretty rampant case of low spotted oak wort in Southern California that have decimated a lot of our oak trees. So we are targeting those trees obviously that are showing signs of mortality or structural defect that will likely be one of those trees that cause our outages.
So all that to say is that we are not trying to target all of those tough, five species for complete removal, nor are we looking to target all of those species with those enhanced clearances? We're looking at every tree, not just the species, looking at all of its site specific environmental factors to determine where we need to remove a tree or get greater clearance.
Speaker 9
Yeah, thanks, Mike. Just wondering, but just back to the specific question, you are looking per at the frequency then as well. Is that correct?
Speaker 18
Yeah. Mind if I jump in on this one? Yeah, we do look at the frequency. We look at both. But again, the goal that we're trying to do here is reduce risk events.
So not only is the frequency important, like if there isn't a volume of those trees, like you mentioned, cypress and century plants, and I know Mike could speak to that we do have an actual plan, additional inspections that are done for cypress and century, given that they are a high risk for our system even, but there's just not enough volume of them to make them one of the targeted species because they only contribute to, I think, maybe like less than half a risk event per year. So, you know, to your point, you're right. Some of the species that we targeted high risk, It's both from the frequency standpoint, like palm, but also from volume standpoint because ultimately, we are trying to reduce risk events to reduce ignitions. And so we need to target the trees, and there's five species that make up the most risk events on our system. To actually do something that's going to reduce risk events on the system, we have to target those species.
But no, you're right in the analysis you did from a pure failure over opportunity. You know, there are some distinctions between, you know, actual, what we would normally call risk in an effort to actually bring down risk events on the system. Does that make sense?
Speaker 17
Sure, I can speak on behalf of SCE here. I want to draw a distinction between what we consider to be at risk and what we consider to be a major contributor to outages, because you're right on point that there is a factor of simple volume of population in this case. So for us, we have identified our outage drivers to be pine, oak, eucalyptus, and palm. So it's a bunch of runners up for fifth place, so we'll keep it at the top four. And really, in fact, I've asked numerous times our arborist team about oaks and pines in particular.
And are we missing something here? Why are they driving outages in the way they are? Do we need to be treating them differently? And we continue to investigate that question, right? It certainly suggests if we have outages from them, there's something to be doing differently, gaining more clearance, for example.
But we don't inherently consider them to be at risk species. We have a top 10 list of at risk species, and on it are palm and eucalyptus. So you can see the alignment between the outages and the at risk in that component. But as I mentioned, oak and pine are not on that list. And so when it comes to marrying up the difference between volume of outages your ability to impact that versus you're at risk, it does require different treatment and analysis to understand how to adjust your program appropriately.
Speaker 15
And I'll just add for PG and E. So the simple answer is yes, we do normalize for the size of the population. We've actually, we've also looked at it by geographical area. So like you've heard some of the other utilities have like their top 10 species or top five, we've actually done the same thing, but we've broken up into smaller regions, right? So our coastal area has a different list of top high risk species than the Sierras do, and so we've broken up into our six regional areas that we have, so basically it's the North Coast, what we call the Central Coast, the Bay Area, the Central Valley, the Sierras, and then the Northern Sierras, or North Valley is what we call it.
Each has their own targeted, or list based on the outage data for their area, as well as the species population in that area. We also actually have even normalized for time of year. We see different kinds of outages being caused during winter, which are less of a factor for wildfire prevention. So snow loading will cause certain trees to fail, so like ponderosa pine, for example, is subject to snow loading failures during the winter months, but during the summer months, typically doesn't have a lot of failures caused. We've also been looking into breaking up and having these top 10 species for different types of causes for outage, right?
So a redwood tree, for example, will shed branches where other species won't. So those are higher risk for an overhang than for a fall in risk. So we've been looking at the data from that perspective too, but I do wanna go back to what SDG and E said, which is part of this for us is driving down risk, and so although we do see that the outages are caused, can be normalized, there still is that risk there. And so I might have a circuit that has thousand Ponderosa pines that have strike potential on it, and while there's a likelihood of only ignition, say once every a thousand trees, well, I can still have a chance for ignition on there. So the risk impact plays a role there.
And how that mostly materializes for us is basically in that tree assessment tool that I mentioned earlier, we put that weighting in there to help inform the decisions. We're looking at some other ways of potentially doing, of expanding that in the future to maybe accelerate or do more targeted approaches, but for right now, are still kind of in the concept phase. But I agree, we do look at the species population and a bunch of other factors, and we don't limit it to just one type of failure, we're looking at it from different types of failures. So fall in, overhang, and then grow in is another factor as well.
Speaker 8
Thank you all. Thank you, Joseph, for your question. Henry Burton from the Public Advocates Office.
Speaker 12
Okay. Thanks. My question is about how the utilities are targeting their enhanced vegetation management programs to address the highest risk circuit segments in their HFGD areas. So your risk models are telling us that a very small number of circuit segments and circuit miles are resulting in a disproportionate amount of the total wildfire risk. Just as an example, think I haven't looked at all the utilities yet, but just as an example, six of Edison's distribution circuits account for 25% of the total wildfire risk on the distribution system.
And 24 circuits out of thousands account for half of the total wildfire risk. So my question is how you're targeting your enhanced vegetation management to the high risk circuit segments. I know PG and E has said that they're setting a goal of doing 80% of that work in 20% of the circuits or circuit segments. But since you're only doing EVM on about 7% of your HFTD miles per year, can you explain why that's an appropriate target?
Speaker 15
So to explain where the 80% so there are other operational challenges that we need to contend with. I think we'd all like it to be as close to 100% as possible, but there's other operational challenges. Permitting is certainly one, you know, customers and other environmental factors, all those weigh in. Some of these are also existing projects that we're carrying over from last year. PG and E's risk model has changed three times each year as we continue to learn more, and so we update the model, and circuits that were in the higher risk categories last year may not be this year.
The other piece is that the next step in our planning is that we also take into consideration work that's already been completed. So a circuit that is high risk, if we've addressed the majority of the circuit, the remaining risk may no longer keep it in that top 20%, so we might move to another area where we can get the most impact, if you will. So that's how we target it. 80% is gonna be, I think, a challenge for us, to be honest. It was kind of a balancing act between wanting to do everything there, and then what's also feasible.
I'll just remind that all of the circuits that we're working, 100% of the circuits are all in HFTDs, right, so they're in the high fire threat. So the reality is that there's that potential to have an ignition anywhere, and so we're trying to go for the places where it's, the impact of the ignition will be greatest, and that's where we see the highest risk, so we're trying to work there first, but it is possible to have ignition anywhere, and we're only working in the high fire threat district area. So even that other 20% is in that high risk area.
Speaker 1
I would just echo, thank you, Henry, for that question. So as we mentioned in our presentation, over half of our overhead, certain files exist within the HFTD. So certainly from a prioritization standpoint, we utilize some of the tools and risk models for most where we may prioritize our enhanced vegetation management work. For example, looking at our, our, the segments in our, based on our VRI modeling, make that decision. I think it's important to kind of state the premise that any part of our service territory that exists within the HFPD, that all it takes is one tree, either shut a branch or fall into the power lines, to cause an ignition.
So though we have to be, judicious in the application of our enhanced vegetation management, we have to be extremely prudent. And one one could argue that we have to on the side of caution. Where we are engaging, utilizing our EVM are on those targeted species. We are not primitive to arbitrary clearances, clearances based on what the specific tree tells us. What is the structure of the tree?
Where is it located? Facing to the power lines? Is there an outage history? What is the BRI? Really, from a prioritization standpoint, we can certainly use some of the risk models.
But we have to be extremely diligent and make sure that we are addressing it, no matter where the fire potential lies. We, for example, we could have a tree that fails outside of the HFPD and, under the right conditions, it could fairly quickly be within the HFPD, it could turn into, you know, collaboration. So we, we're going to stand by our, our operations of, of being, extremely, prudent, but also extremely thorough in where we are applying these greater clearances. Three, let the data inform us on where we're gonna do it. We would argue that the mitigation efforts that they are taking, and our data is showing that, has resulted in a reduction of outages and thus
Speaker 17
Yeah, in terms of SCE, I think, a lot of good points there that I don't need to repeat. A couple points I would make have to do with specifically your question and the reference to certain distribution circuits being the majority driver of risk. And you see that come in, in terms of risk informed planning with regard to the hazard tree program, which really seeks to work down and evaluate the trees along those circuits in just that fashion. There is a component for operationalizing the work and the need to not be hither and thither all over the territory with a segment of a line here and a segment of line there because then you quickly spiral your costs out of control and really find it difficult to maintain momentum in the work. But by and large, the approach is to work down in terms of risk severity.
And then the other nuance there relates to things like our line clearing, where we're doing enhanced clearances, right, which is that 12 foot clearance at time of trim that has been discussed extensively over the past year plus. And really, we do that in all locations in the high fire risk areas. And that's because it it not only reduces the wildfire potential, but it also, is supportive of the compliance obligation to not have the tree encroach on the minimum clearance distance. And so there's a component of risk informed points there that, again, I'm not trying to repeat or belabor points, but there may be variations in trees in terms of their growth cycle and our ability to obtain a lesser distance in some cases. But we anchor on a standard in order to provide consistency, in the direction to our workforce so that they know what's expected.
And because we're going to those locations every single year and evaluating the condition of individual trees, the risk associated with that particular circuit today in the world as we have it is less consequential. Now as we mature our risk modeling, we may move to a place where that is more informed and that's something we're trying to get to. But right now, we're not quite there. And so that point is not as relevant for that particular program.
Speaker 15
I'd also like to just echo something Melanie said is, so PG and E as well, our high fire threat districts get patrolled twice a year, every year. So we are still making sure that we're working and doing our radio clearance, removing dead and dying trees and other trees that we think are defective and need to go. When the enhanced EBM program, Enhanced Vegetation Management program gets there, that program is going above and beyond the requirements and attempting to get a greater clearance, establish a bigger right of way, inventory all the trees. And so, it's not that the rest of the system isn't getting attention from us, It's just not getting that enhanced program yet because to your point, we're only able to do so many of those miles annually, but we do still patrol the rest of the system twice a year, every year.
Speaker 12
I have other questions, but it looks like we're out of time and I'm sure other people have other questions as well.
Speaker 8
We are out of time and I would like to elevate one more person before we are done for the day. That is alright. Megan from the Joint Local Governments Coalition, I see your question in the chat, and if you would like to unmute yourself and ask your question, you are welcome to.
Speaker 20
Yes. Thank you. This is Megan Samaji for the Joint Local Government Coalition. And this question is primarily for PG and E, although, there there's some things that SCE and SDG and E could speak to as well. I wanna follow-up about the fire retardant again.
Now your wildfire mitigation plan speaks to getting pesticide use permits and, undertaking CEQA and NEPA review on state and federal lands for your rights of way where you're looking to potentially, deploy this fire retardant. But for work on private property, you only mention a lands right assessment. And so in light of the fact that in your service territory, have significant private agricultural, property that your equipment runs through. You have coastal property, in the high fire threat area, and you have a large number of customers who rely on the groundwater table for their well water. I'd like to know what PG and E's plan is at a programmatic level for comprehensive environmental review and permitting for this fire retardant?
Speaker 15
No, thank you, it's a great question, and I just wanna clarify, you said pesticide, it should be herbicide, hopefully it says that in the Wildfire Mitigation Plan, it should say herbicide, not pesticide.
Speaker 15
is why we haven't started yet. We are currently going through an environmental review, we've been working with several suppliers of flame retardant to understand what its environmental impacts are, and we are still going through that practice. Certainly one thing that initially came out is that some of these fire retardants, if not all, do have a very negative impact on the water streams, and so that's certainly something that we've already made note of, but that environmental review is still undergoing, and we certainly don't wanna do anything that's going to damage the environment and make things worse. I personally live off of a well water system, so I definitely can relate to that as well. As for what the plan is, we're still awaiting the results of that review and understanding what the impacts will be.
We're looking at products that are used already in the process of fire prevention, so a lot of the samples, or at least one of them is actually the same product that CAL FIRE uses in their fire protection activities. But I don't have the details to answer your question, because we're still trying to understand that, and we're not going to apply these products until we have that understood, and make sure that it's not causing more harm than it's preventing.
Speaker 20
Thank you. And just a very brief follow-up. Once you do actually have the environmental work done and have a better understanding of what this might do to the environment, do you plan to make that work or those studies publicly available to the stakeholders at the commission and also directly to the impacted cities, counties and state agencies?
Speaker 15
Barring some unknown legal requirements that I'm not aware of, I would say yes. I mean, we make, I think, all of that information available, but barring some legal requirements that I'm not personally aware of right now, I would say the answer to that is yes. I'm sure, I know we've already talked about internally what the outreach would look like in terms of talking with communities and allowing communities to weigh in on this, but before we go into that step, we wanted to have the results. I I'd like to think that if we find that this stuff is damaging to the environment, we're just not gonna use it. So I don't, I know we'll work with the local communities, and I'm sure we'll make all of our information available to them as we progress through this.
We want people to feel confident that this is a good choice that we're making, and that it actually helps to prevent wildfires while also helping to keep this state as beautiful and healthy as it is.
Speaker 1
And if I could, I'll just speak to SDG and E's efforts regard to environmental reviews. So, as mentioned earlier, our pilot project with the use of fire turd, area was about 25 acres. It go through, a review from our safety department as well as our environmental department. And, anytime we do any activities, especially new activities, our environmental s- the environmental services department reviews that to make sure that also within guidelines of our natural communities conservation plan, which is essentially a landscape level habitat conservation plan that SDG has with the wildlife agency. This particular pilot, we did limit the scope to specific areas where we do not see, environmental damage or negative impact, but it did go through that environmental review in advance of us actually in that product too.
And that can be made available.
Speaker 8
Thank you for your question, Megan. It seems that Melanie from SCE is having computer issues. And so I think with that, we will conclude for the day as we are at time. Thank you to all of our panelists for answering various questions from our stakeholders and from the public. There are some final words from Melissa Spencer to send us off for the day.
Speaker 2
I'm not sure if I'm showing up. Okay. There we go. Yeah, I don't have much more to add than that. Just thank you to everyone who participated today, for the utilities, for preparing their presentations, for the panelists that joined and asked the questions.
We have tried to capture the questions in the chat that have not been answered, mainly from this morning, so that we can follow-up on those. For the stakeholders who did not get all of your questions answered, which I know was the case for just about all of you, feel free to follow-up with those questions and data requests to the utilities. And other than that, we are back on at 09:30 again tomorrow morning, and we'll use the same procedures as we did today. So thank you for everyone for showing up, and we will see you tomorrow. Thank you.
Ask anything about this video — I analyse its transcript to answer.